Accepted Manuscript

Title: Metformin reduces all-cause mortality and diseases ofageing independent of its effect on diabetes control: asystematic review and meta-analysis

Authors: Jared M. Campbell, Susan M. Bellman, Matthew D.Stephenson, Karolina Lisy.

PII: S1568-1637(17)30147-2DOI: http://dx.doi.org/doi:10.1016/j.arr.2017.08.003Reference: ARR 782

To appear in: Ageing Research Reviews

Received date: 6-7-2017Revised date: 3-8-2017Accepted date: 3-8-2017

Please cite this article as: Campbell, Jared M., Bellman, Susan M., Stephenson,Matthew D., Lisy., Karolina, Metformin reduces all-cause mortality and diseases ofageing independent of its effect on diabetes control: a systematic review and meta-analysis.Ageing Research Reviews http://dx.doi.org/10.1016/j.arr.2017.08.003

This is a PDF file of an unedited manuscript that has been accepted for publication.As a service to our customers we are providing this early version of the manuscript.The manuscript will undergo copyediting, typesetting, and review of the resulting proofbefore it is published in its final form. Please note that during the production processerrors may be discovered which could affect the content, and all legal disclaimers thatapply to the journal pertain.

1

Metformin reduces all-cause mortality and diseases of

ageing independent of its effect on diabetes control: a

systematic review and meta-analysis

Jared M. Campbell,1,2 Susan M. Bellman,1 Matthew D. Stephenson,1 Karolina Lisy.3

1 The Joanna Briggs Institute, The University of Adelaide, Adelaide, South Australia

2 Centre for Nanoscale BioPhotonics, Macquarie University, Sydney, New South Wales

3 Peter MacCallum Cancer Centre, Melbourne, Victoria

Corresponding author:

Jared M. Campbell

55 King William Street, North Adelaide, 5006

Jared.campbell@mq.edu.au

+61 (08) 83132279

Highlights

Diabetics on metformin have lower morality than non-diabetics and other diabetics.

Diabetics on metformin have less cancer than non-diabetics and other diabetics.

Diabetics on metformin have less cardiovascular disease than other diabetics.

Metformin appears to extend health and life spans independent of its effect on

diabetes.

Metformin may be able to extend health and lifespans in the general population.

Abstract

This systematic review investigated whether the insulin sensitiser metformin has a

geroprotective effect in humans. Pubmed and Embase were searched along with databases

of unpublished studies. Eligible research investigated the effect of metformin on all-cause

mortality or diseases of ageing relative to non-diabetic populations or diabetics receiving

other therapies with adjustment for disease control achieved. Overall, 260 full-texts were

reviewed and 53 met the inclusion criteria. Diabetics taking metformin had significantly

lower all-cause mortality than non-diabetics (hazard ratio (HR)= 0.93, 95%CI 0.88-0.99), as

did diabetics taking metformin compared to diabetics receiving non-metformin therapies

(HR= 0.72, 95%CI 0.65-0.80), insulin (HR=0.68, 95%CI 0.63-0.75) or sulphonylurea (HR= 0.80,

95%CI 0.66-0.97). Metformin users also had reduced cancer compared to non-diabetics (rate

ratio= 0.94, 95%CI 0.92-0.97) and cardiovascular disease (CVD) compared to diabetics

receiving non-metformin therapies (HR= 0.76, 95%CI 0.66-0.87) or insulin (HR= 0.78, 95%CI

2

0.73-0.83). Differences in baseline characteristics were observed which had the potential to

bias findings, although statistical adjustments were made. The apparent reductions in all-

cause mortality and diseases of ageing associated with metformin use suggest that

metformin could be extending life and healthspans by acting as a geroprotective agent.

Keywords: metformin; aging; insulin sensitizer; lifespan; longevity; geroprotection

1.0 Introduction

Ageing is characterised by progressive damage at a cellular and organ level as prevention

and repair mechanisms begin to break down (Finkel and Holbrook, 2000; Garinis et al., 2008;

Mitchell, 2008). Diseases of ageing – such as cardiovascular disease (CVD), kidney failure and

cancer – are the result of the accumulation of this damage and ultimately responsible for

organisms dying of old age. With the ageing population of the developed world the

extension of healthspan – the period that an individual is functional and free of chronic

diseases (Barzilai et al., 2016) – has been identified as an important goal both for the

minimisation of human suffering and to enable healthcare systems to cope.

One drug which has been subjected to significant research as a geroprotective factor is the

insulin sensitiser metformin. Metformin has been shown to extend the lifespans of model

organisms (including mice and C. elegans (Anisimov et al., 2011; Anisimov et al., 2010a;

Cabreiro et al., 2013)) and a number of potential mechanisms for its effects have been

discussed, including decreased insulin and IGF-1 signalling (Liu et al., 2011), inhibition of

mTOR (Kickstein et al., 2010; Nair et al., 2014), reducing the levels of reactive oxygen species

(ROS)(Batandier et al., 2006; Zheng et al., 2012), lowering inflammation (Saisho, 2015),

reducing DNA damage (Algire et al., 2012; Na et al., 2013), and the activation of AMP-

activated protein kinase (AMPK)(Zhou et al., 2001). Its effect on AMPK has received

particular attention as it models the intracellular mechanisms of caloric restriction (Gillespie

et al., 2016; Lee and Min, 2013) – another intervention which has been found to be capable

of extending the lifespans of animal models – but which is not feasible for widespread

implementation in humans (Roth and Ingram, 2016). This, among other factors, has led to

metformin being described as a caloric restriction mimetic (CRM) (Gillespie et al., 2016; Lee

and Min, 2013; Roth and Ingram, 2016).

One aspect of metformin which makes it particularly promising for development as a

geroprotective agent is that it is already being widely used in humans for a different

purpose. As an insulin sensitiser metformin is a first line therapy for diabetes (Nathan et al.,

3

2006). Through this use its potential for adverse effects and contraindications have been

well characterised, and it has been found to have a broad safety profile (Rojas and Gomes,

2013) which would make it significantly quicker and easier to implement as an intervention

for ageing than a previously unutilised drug.

Due to the widespread application of metformin for the management of diabetes, a large

amount of data has already been collected on its effects on mortality and diseases of ageing.

However, extrapolation of this research evidence to the general population is invariably

confounded by the fact that the people who are currently receiving metformin have

diabetes and are taking it for the treatment thereof. As such, any benefits seen could be due

to improvements in diabetes control. Two analytic strategies have been attempted to

overcome this limitation and gain some insight as to whether metformin could be used to

improve the healthspan of the general population.

The first strategy is the comparison of the outcomes of diabetics taking metformin to the

general or non-diabetic population (Bannister et al., 2014). This approach reasons that as

diabetics are on average less healthy than non-diabetics, superior outcomes in the

metformin/diabetic group should be a consequence of beneficial effects of metformin

overcoming their generally worse health status. A weakness of this strategy is that people

taking metformin could be observed to have worse or the same outcomes as non-diabetics

in the presence of an actual beneficial effect of metformin due to it being masked by the

detrimental effects of diabetes. The second strategy is to compare diabetics taking

metformin to diabetics managing their diabetes through other means while statistically

controlling for the effects of the different therapies on diabetes management (i.e. HbA1c or

diabetes related doctor visits (Lin et al., 2015)). This strategy reasons that any difference in

outcomes seen after adjusting for the relative effects on disease management can be

attributed to metformin’s activity beyond diabetes control, and are therefore generalisable.

This strategy’s weaknesses are that statistical adjustment is an imperfect process – it is

always possible that residual confounding remains that could cause any differences

observed – and that other antidiabetic drugs could cause harms which in comparison would

appear to be beneficial effects by metformin.

However, although additional controlled, experimental research will need to be carried out,

there exists a wealth of observational data that can offer evidence on whether metformin

could be applied as a geroprotective agent in humans. As such, this systematic review and

meta-analysis has been undertaken to identify and synthesise all studies where the effects

4

of metformin on all-cause mortality or diseases of ageing have been compared to the

general or non-diabetic population or to diabetics managing their diabetes through other

means with adjustment for disease control. Numerous systematic reviews have been carried

out which nominally address this review’s outcomes of interest, however those which

investigate the association between metformin and all-cause mortality do so in the context

of severe diseases in addition to diabetes (usually various forms of cancer) (Coyle et al.,

2016; Gash et al., 2017; Li et al., 2017; Meng et al., 2017; Tang et al., 2017; Zhou et al.,

2017a) and their results therefore cannot be generalised , while those that investigate the

incidence of diseases of ageing do so without attempting to adjust for diabetes control (Liu

et al., 2017; Ma et al., 2017; Pladevall et al., 2016; Tang et al., 2017; Zhou et al., 2017b). As

such, the question of whether metformin could act as a geroprotective agent in humans has

yet to be addressed in a systematic review.

2.0 Methods

2.1 Objective

This systematic review aimed to identify and synthesise all research on the effect of

metformin on all-cause mortality and diseases of ageing which had the potential to give

evidence on whether it could be used as a geroprotective factor to extend life and

healthspan in the general population. It followed an a priori protocol pre-registered with

PROSPERO (CRD42016036098).

2.2 Inclusion criteria

2.2.1 Population

Studies on all-cause mortality were eligible for inclusion if they reported on adults with a

minimum age of 40 years or a mean age of 50 years. This age restriction was applied to

increase the proportion of deaths which would be attributable to age and was not applied to

studies on diseases of ageing. Studies were included where metformin was being given to

patients for the treatment of a particular disease (diabetes in all cases although this was not

an inclusion criterion), however where populations were specifically selected to include a

second serious illness (i.e. patients with cancer who were also being treated for diabetes),

studies were not eligible for inclusion.

2.2.2 Exposure

5

Studies were included where a group of patients had been treated with metformin of any

dosage taken on an ongoing basis.

2.2.3 Comparator

Any controls were eligible for inclusion provided they were not receiving metformin. This

included members of the general or non-diabetic population as well as people who were

receiving other therapies for diabetes. In the latter case, studies were only included if data

was adjusted for the effect of diabetes on disease control (i.e. HbA1c levels or number/rate

of diabetes related healthcare visits). Where patients entered into studies were using

metformin on an ongoing basis their data was included if diabetes control at baseline was

adjusted for, as this reflected how well their diabetes was being managed by metformin.

However, if patients were enrolled with incident metformin use and the only adjustment for

disease control was made at baseline, the study was excluded as this data would not reflect

the effect of metformin.

2.2.4 Outcomes

The two primary outcomes were all-cause mortality and the incidence, onset or prevalence

of diseases of ageing. Studies on disease specific mortality were not eligible for inclusion.

Any diseases of ageing were eligible for inclusion with the exception of diabetes as any

effect metformin had on diabetes incidence could more likely be attributed to it directly

treating pre-diabetes as an insulin sensitiser than extending healthspan. Diseases ultimately

included were cancer (including breast, lung, colorectal, pancreatic, prostate, oesophageal,

thyroid, renal, hepatocellular, and head and neck cancer), CVD (including stroke, myocardial

infarction, heart failure, coronary heart disease, macrovascular morbidity and ventricular

dysfunction), kidney failure, fracture, open angle glaucoma, cognitive impairment, and

carpal tunnel syndrome.

2.2.5 Studies

Both experimental and observational studies (including cohort, case-control and analytical

cross-sectional studies) were eligible for inclusion in this systematic review. Case report and

case series studies were ineligible.

2.3 Search strategy

The search for published literature included Pubmed and Embase, while the search for

unpublished studies included the International Clinical Trials Registry Platform, ProQuest

6

Dissertations and Theses Global, and OpenThesis. No date restrictions were applied,

however only English language articles were eligible for inclusion. The search was carried out

in March 2016. Full details of the structure and content of each search undertaken are

reported in supplementary material 1. Titles and abstracts were screened followed by the

retrieval of potentially relevant full-texts. These were then carefully compared to the

inclusion criteria to identify eligible studies. The reference lists of included full-texts were

then reviewed, however no additional studies that met the inclusion criteria were identified.

2.4 Appraisal and extraction

Duplicate critical appraisal was carried out using the standardised Joanna Briggs Institute

(JBI) appraisal checklists for experimental, cohort, case control and cross-sectional studies

(The Joanna Briggs Institute, 2014) by independent reviewers working separately, with

consensus reached through discussion in all cases. Data was extracted for all studies using an

extraction template which included fields for; study methodology, study design, inclusion

criteria, data source, country, disease control, exposure, comparator, sample size/events,

follow up, outcome(s), outcome data, and statistical adjustments. Where additional data or

clarification was needed, attempts were made to contact corresponding authors by email.

2.5 Data synthesis

Where possible, data was pooled through meta-analysis. The inverse variance method with

a random effects model from RevMan was used, unless otherwise specified. Data included in

the meta-analyses was always from the analyses with the greatest adjustment. Where

summary measures differed, outcomes were converted to hazard ratios (HRs) using the

equations described in Tierney et al. 2007 (Tierney et al., 2007). Data were analysed

separately based on comparators utilised, and heterogeneity was assessed using the Chi2

and I2 tests. Two sets of sensitivity analyses were carried out where poor quality studies

(<70% on critical appraisal) were removed from meta-analyses, and where studies that did

not adjust for duration of diabetes and/or comorbidities were removed. Analysis of

publication bias through Funnel plots and Egger’s test was planned if sufficient studies (at

least 10) were found for any outcomes, however none were. Where statistical pooling was

not possible data is presented narratively.

7

3.0 Results

The initial search identified a total of 21,027 studies, which was reduced to 19,408 following

the removal of duplicates. Following title/abstract screening, 260 studies were included for

full-text review. Of these, 207 were excluded and 53 were included. Full search details and

reasons for exclusion are given in Fig. 1. Overall, 13 included studies reported data on all-

cause mortality while 49 reported on diseases of ageing.

3.1 All-cause Mortality

Of the studies that investigated the effect of metformin on all-cause mortality, four

compared diabetic patients being treated with metformin to the general population or non-

diabetic patients (Bannister et al., 2014; Berard et al., 2011; Bo et al., 2012; Claesen et al.,

2016), while nine compared diabetic patients being treated with metformin to diabetic

patients managing their diabetes through other means. Subgroups within the latter group

included insulin therapy (Ekstrom et al., 2012; Ghotbi et al., 2013), diet interventions (Bo et

al., 2012; Sullivan et al., 2011), sulphonylurea therapy (Evans et al., 2006; Kahler et al., 2007;

Sullivan et al., 2011; Wang et al., 2014), and non-metformin therapies (studies assigned to

this group included all patients receiving any other oral hypoglycemic agent, but may have

also included those receiving insulin, diet, or no therapy as well) (Ekstrom et al., 2012;

Ghotbi et al., 2013; Gosmanova et al., 2008; Libby et al., 2009; Pratipanawatr et al., 2010).

Full details of interventions, controls and other study characteristics are included in table 1.

For the metformin versus non-diabetic comparison three of the cohorts were matched by

age (Bannister et al., 2014; Bo et al., 2012; Claesen et al., 2016) while in the fourth (Berard et

al., 2011) metformin users were older (mean±SD age 56.2±6.6) than non-diabetics

(50.1±8.3); a disparity which could be expected to bias the results towards a finding of

metformin being associated with increased mortality, although age was adjusted for. In the

majority of cases when metformin users were compared to other diabetics, metformin users

were younger (Ekstrom et al., 2012; Evans et al., 2006; Ghotbi et al., 2013; Gosmanova et al.,

2008; Kahler et al., 2007; Sullivan et al., 2011; Wang et al., 2014). Differences between mean

ages ranged from 0.4 years (Gosmanova et al., 2008) to 5.9 years (Ekstrom et al., 2012). In

one study the metformin group was older than the control by <1 year (Bo et al., 2012), and

in two mean ages were not reported (Libby et al., 2009; Pratipanawatr et al., 2010), although

Libby et al. did match by year of diagnosis so age is likely to have been close. The trend of

metformin users being younger than other diabetics has the potential to bias results towards

a finding of decreased mortality, however all studies adjusted for age, with the exception of

8

Bo et al. The appraisal of study quality found that overall the studies were conducted

rigorously, with notable deficiencies being that none reported similar characteristics in their

respective populations at baseline and less than half had adequate follow up times for our

observations of interest (defined as ≥5 years) with thorough reporting on reasons for loss to

follow up (Supplementary material 2).

9

Meta-analysis of the four studies that compared people managing their diabetes with

metformin to non-diabetics or the general population (Bannister et al., 2014; Berard et al.,

2011; Bo et al., 2012; Claesen et al., 2016) found that people taking metformin had

significantly lower mortality compared to those who were not (Fig. 2 HR= 0.93, 95%CI 0.88

to 0.99, p=0.03). One included study was an outlier in that it reported a non-significant

increase in mortality for people taking metformin (Berard et al., 2011), however it was by far

the smallest study, with just nine events in the metformin group. Hazard ratios (HRs) used in

the meta-analysis represent adjusted values with two exceptions where the HR could only

be calculated for crude data (Bannister et al., 2014) and where only crude analyses were

performed for this data (Bo et al., 2012). Sensitivity analysis could not be carried out as no

studies were of poor quality and only one study adjusted for diabetes duration and/or

comorbidity.

The finding that metformin reduces mortality was supported by analyses that compared

diabetic patients receiving metformin to diabetic patients receiving other therapies (Fig. 3),

where diabetes control had been adjusted for.

The majority of subgroups, including studies with non-metformin controls (HR= 0.72, 95%CI

0.65 to 0.80, p<0.00001), insulin treated controls (HR=0.68, 95%CI 0.63 to 0.75, p<0.00001),

and sulphonylurea treated controls (HR= 0.80, 95%CI 0.66 to 0.97, p= 0.02), found that even

after adjusting for metformin’s effect on diabetes control, diabetics taking it had lower all-

cause mortality than diabetics who were not. The one exception was the subgroup where

the controls managed their diabetes through diet alone. In this case one study (Bo et al.,

2012) found a significant benefit to all-cause mortality for metformin and the other study

(Sullivan et al., 2011) found a non-significant disadvantage resulting in an overall non-

significant finding (HR=0.91, 95%CI 0.68 to 1.22, p= 0.53). An overall meta-analysis could not

be performed as the same studies were included across multiple subgroups where they

10

presented their metformin data relative to different controls. However, when the duplicate

data sets were excluded (removing those with the strongest positive findings: Ekstrom Non-

Metformin, Ghotbi Non-Metformin and Sullivan Sulphonylurea), diabetics taking metformin

had significantly lower mortality compared to controls after adjusting for disease control

(HR= 0.75, 95%CI 0.69 to 0.82, p<0.00001). Sensitivity analysis for the non-metformin

control subgroup where two studies which did not adjust for duration of diabetes and/or

comorbidity were excluded (Gosmanova et al., 2008; Libby et al., 2009) did not affect

results. Sensitivity analysis for study quality could not be performed as no studies were poor

quality.

3.2 Diseases of ageing

The main diseases of ageing investigated in the included studies were cancer (Supp table 1)

and cardiovascular disease (CVD, Supp table 2). However, other age related conditions were

investigated in a smaller number of studies including renal failure, fracture, open angle

glaucoma, cognitive impairment and carpal tunnel syndrome (Supp table 3).

3.2.1 Cancer

Overall, 25 studies investigated the association between metformin use and cancer (Supp

table 1). Where people taking metformin were compared to the general or non-diabetic

population the majority of studies did not report the ages of these groups. The two that did

found that people taking metformin were significantly older (Andersson et al., 2012; But et

al., 2014), potentially biasing the results to increased risk of cancer in metformin users

although age was adjusted for in both analyses. Amongst the remainder of studies half were

case-controls which matched for age and are therefore unlikely to have meaningful hidden

bias (Becker et al., 2014, 2016; Bodmer et al., 2012a; Bodmer et al., 2012b; Lu et al., 2015;

Walker et al., 2015) and the remainder adjusted for age (Baur et al., 2011; Becker et al.,

2013; Nordström et al., 2015; Tseng, 2011, 2012a, b). For people taking metformin

compared to diabetics controlling their diabetes through other means, only one study

reported relative ages for the comparison (metformin users were younger biasing results

towards a lower risk of cancer although age was adjusted for (Goossens et al., 2015)). Nine

studies were case-controls which matched for age and therefore have limited risk of bias in

this regard (Azoulay et al., 2011; Becker et al., 2013; Bodmer et al., 2010; Bosco et al., 2011;

Chen et al., 2013; Mazzone et al., 2012; Sehdev et al., 2015; Smiechowski et al., 2013a;

11

Smiechowski et al., 2013b), while two cohort studies adjusted for age but did not report it

(Libby et al., 2009; Redaniel et al., 2012). Critical appraisal showed that cohort studies

compared groups with significant differences at baseline, tended to lack adequate follow up

and did not report on reasons for loss to follow up with similar deficiencies observed in case-

control studies (supplementary material 2). Of the included studies, three investigated the

incidence of any cancer for people taking metformin for diabetes control compared to non-

diabetics. Two of these studies could be combined in meta-analysis (Fig. 4A, (Andersson et

al., 2012; But et al., 2014)) and showed that diabetics taking metformin had significantly

lower cancer incidence (Rate ratio= 0.94, 95%CI 0.92 to 0.97, p=0.0003). In this case rate

ratio was used in the meta-analysis rather than HR as both studies used it as the summary

measure to report their data. It should be noted that although there was essentially no

heterogeneity between the two studies, the results of the meta-analysis drew almost

entirely on the findings of (Andersson et al., 2012) which received 99% of the weight. The

third study (Baur et al., 2011) was cross-sectional and found a non-significant decrease in

the prevalence of cancer in diabetics receiving metformin monotherapy compared to the

non-diabetic population, and a non-significant increase in diabetics receiving metformin as a

combination therapy. A fourth study investigated the effect of metformin on cancer

incidence but compared it to matched diabetics receiving any other oral hypoglycaemic

agent rather than non-diabetics (Libby et al., 2009). It reported a significant reduction for

metformin (HR= 0.63, 95%CI 0.53 to 0.75, p<0.05), supporting the findings of the meta-

analysis.

Three studies investigated the effect of taking metformin on the development of pancreatic

cancer in people with diabetes compared to the general (Bodmer et al., 2012b; Lu et al.,

2015) or non-diabetic population (Walker et al., 2015). However, findings were mixed and

the result of the meta-analysis was non-significant (Fig. 4B, all studies reported their

outcomes as odds ratios (ORs), therefore this was the summary measure used in the meta-

analysis). Sensitivity analysis could not be carried out as no studies were of poor quality and

only one study adjusted for duration of diabetes and/or comorbidity.

12

The effect of metformin on the incidence of colorectal cancer was investigated in four

studies, three of which were similar enough to be combined in meta-analysis (Libby et al.,

2009; Sehdev et al., 2015; Smiechowski et al., 2013a). These studies compared the incidence

of colorectal cancer in diabetic patients taking metformin to non-metformin diabetic

controls. Although a trend for metformin use being associated with a reduced incidence of

colorectal cancer was found, the results were ultimately non-significant (Fig. 5A, HR= 0.85,

95%CI 0.72 to 1.02, p=0.08). As heterogeneity was non-significant (Chi2= 0.15, I2= 48%) and

all studies suggested a benefit, a meta-analysis using the fixed effects model was trialled. A

significant reduction in colorectal cancer incidence for the metformin group was found (HR=

0.91, 95%CI 0.84 to 0.98, p=0.009). Sensitivity analysis where one study which did not adjust

for duration of diabetes and/or comorbidity was excluded (Libby et al., 2009) did not have a

large effect on the estimate but resulted in a significant finding by the random effects model

(HR= 0.92, 95%CI 0.85 to 0.99, p=0.02). Sensitivity analysis for study quality could not be

performed as no studies were poor quality. An additional study not included in the meta-

analysis (Tseng, 2012a) compared the incidence of colorectal cancer in diabetics taking

metformin to the general population and found that metformin was associated with a

significant reduction (relative risk= 0.73, 95%CI 0.58 to 0.92, p=0.008), supporting the

conclusion that metformin may reduce colorectal cancer incidence.

Breast cancer incidence was investigated in four studies, three of which compared the

incidence in diabetic patients taking metformin to diabetic patients who were not taking

metformin (Bodmer et al., 2010; Bosco et al., 2011; Libby et al., 2009). These were combined

in meta-analysis Fig. 5B which showed a significant reduction in the incidence of breast

cancer (HR= 0.71, 95%CI 0.54 to 0.92, p=0.01). Sensitivity analysis where one study which did

not adjust for duration of diabetes and/or comorbidity was excluded (Libby et al., 2009) did

not have a large effect on the estimate but resulted in a non-significant finding being made

(HR= 0.72, 95%CI 0.50 to 1.04, p=0.08). Sensitivity analysis for study quality could not be

performed as no studies were poor quality. The fourth study compared the incidence of

breast cancer in diabetics being treated with metformin to diabetics being treated with

sulphonylurea (Redaniel et al., 2012) and found a non-significant increase in breast cancer

for metformin. Interestingly, they also compared people taking metformin in combination

with sulphonylurea to people taking sulphonylurea alone and found a significant reduction in

the incidence of breast cancer in the people treated with metformin (HR= 0.66, 95%CI 0.50

to 0.88, p<0.05).

13

The effect of metformin on the incidence of lung cancer was investigated in four studies.

Three studies compared the incidence of lung cancer in diabetic patients treated with

metformin to diabetic patients not treated with metformin (Libby et al., 2009; Mazzone et

al., 2012; Smiechowski et al., 2013b) and could be combined in meta-analysis Fig. 5C. The

combined results showed that taking metformin was associated with a decreased incidence

of lung cancer in diabetic patients (HR= 0.80, 95%CI 0.65 to 0.98, p=0.03). Sensitivity analysis

could not be carried out as no studies were of poor quality and only one study adjusted for

duration of diabetes and/or comorbidity. The fourth study (Bodmer et al., 2012a) compared

diabetics taking metformin to the general population. It did not find any significant effects

for short term (1-14 prescriptions) or long term (≥40 prescription) metformin use, however

medium term (15-39 prescriptions) metformin use was contradictorily associated with a

significant increase in the odds of developing lung cancer (OR= 1.24, 95%CI 1.03 to 1.50,

p<0.05).

Additional cancers that were investigated included prostate, bladder, thyroid, renal, head

and neck, oesophageal, and hepatocellular cancer. Neither of the two studies on prostate

cancer found any significant association with metformin use (Azoulay et al., 2011;

Nordström et al., 2015). Nor did the studies on bladder cancer (Goossens et al., 2015; Tseng,

2011), thyroid cancer (Tseng, 2012b), renal cell carcinoma (Becker et al., 2016), head and

neck cancer (Becker et al., 2014), or oesophageal cancer (Becker et al., 2013). However,

metformin users had significantly lower odds of developing hepatocellular cancer compared

to other diabetics (OR= 0.79, 95%CI 0.75 to 0.85, p<0.0001).

3.2.2 Cardiovascular disease

Fifteen studies were found that investigated the effect of metformin on the development of

CVDs (Supp table 2). Critical appraisal showed a similar pattern in quality as observed for

studies on all-cause mortality and cancer – groups had significant differences at baseline,

follow-up was frequently shorter than desirable to observe the onset of disease, and loss to

follow up was inadequately described (supplementary material 2). Additionally, over half of

14

the cohort studies included participants who already had some form of CVD at baseline and

half of the case-control studies did not adequately control for potential confounding factors.

No studies compared people taking metformin for diabetes management to non-diabetics or

the general population. For those that compared people taking metformin to people

controlling their diabetes through other means, six included analyses where the group

receiving metformin was younger than those who were not (Ekstrom et al., 2012; Evans et

al., 2006; Ghotbi et al., 2013; McAlister et al., 2008; Sullivan et al., 2011) (the difference

ranged from 1.1 years (Sullivan et al., 2011) to 5.9 years (Ekstrom et al., 2012), while in two

studies people receiving metformin were younger than comparators (2.8 years (Nichols et

al., 2005) and 5 years (Kooy et al., 2009)). Three further studies were case-controls which

matched for age (Floyd et al., 2016; Hartung et al., 2005; Koro et al., 2005) while four did not

report on relative ages but did adjust for age in their analyses (Gejl et al., 2015; Giorda et al.,

2011; Jansson et al., 2014; Peters et al., 2013). Five studies comparing the incidence of any

CVD in diabetics taking metformin to diabetics not receiving metformin could be combined

in meta-analysis (Ekstrom et al., 2012; Gejl et al., 2015; Ghotbi et al., 2013; Jansson et al.,

2014; Peters et al., 2013). This showed a significant reduction in CVD for people taking

metformin (Fig. 6, HR= 0.76, 95%CI 0.66 to 0.87, p<0.0001). Sensitivity analysis where

studies of low quality were excluded from meta-analysis did not have a large impact on the

estimate (HR= 0.73, 95%CI 0.53 to 1.00, p=0.05). No studies did not adjusted for duration of

diabetes and/or comorbidity. The finding of a reduction was supported by two studies that

compared the incidence of any CVD between diabetics being treated with metformin to

diabetics treated with insulin (Ekstrom et al., 2012; Ghotbi et al., 2013), where a significant

decrease in incidence with metformin use was found (Fig.6, HR= 0.78, 95%CI 0.73 to 0.83,

p<0.00001). However, the meta-analysis for sulphonylurea controls, which also included two

studies (Evans et al., 2006; Sullivan et al., 2011), did not show a significant effect (Fig. 6). One

study was found that investigated the effect of metformin on incidence of CVD compared to

diabetics managing their diabetes through diet and showed a non-significant increase in the

metformin group (Sullivan et al., 2011). An overall meta-analysis was prevented as the

subgroups contained data from the same participants. However, when duplications were

removed (again excluding the data that showed the strongest positive effect: (Ekstrom Non-

metformin, Ghotbi Non Metformin, Sullivan Diet), the result was a significant effect for

metformin reducing the incidence of CVD (HR= 0.83, 95%CI 0.73 to 0.94, p=0.004).

15

Three studies investigated the effect of metformin on the incidence of stroke. Two

compared the incidence of stroke in metformin users to other diabetics not using metformin

(Floyd et al., 2016; Jansson et al., 2014). Meta-analysis of their findings (Fig. 7A) showed a

significant reduction in stroke for metformin users (HR= 0.70, 95%CI 0.53 to 0.93, p=0.01).

The third study (Sullivan et al., 2011) compared metformin use to diabetics managing their

diabetes through diet alone or with sulphonylurea, and found a non-significant increase in

the incidence of stroke with metformin use in both cases.

Myocardial infarction incidence was examined in just two studies (Floyd et al., 2016; Jansson

et al., 2014) and although (Jansson et al., 2014) showed a strong significant effect in favour

of metformin, the result of the meta-analysis was a non-significant decrease in myocardial

infarction for metformin users compared to diabetics not using metformin (Fig. 7B, HR=

0.63, 95%CI 0.28 to 1.42, p=0.27).

Heart failure was investigated in four studies, however due to differences in study design

and outcomes reported no meta-analysis could be performed. One study (Hartung et al.,

2005) compared metformin users to other oral hypoglycaemic agent users with regards to

hospitalisations resulting from heart failure, but found no significant effect. Another study

(Koro et al., 2005) found non-significant increases for metformin on incidence of congestive

heart failure compared to diabetics with no drug exposure and diabetics treated with

sulphonylurea. Another study (Nichols et al., 2005) showed a non-significant decrease in

congestive heart failure for diabetics treated with metformin compared to those treated

with sulphonylurea. It also found a significant decrease in heart failure with metformin

compared to insulin use (rate ratio= 0.39, 95%CI 0.19 to 0.80, p<0.05). The final study

(McAlister et al., 2008) compared the incidence of heart failure in metformin treated

patients to those treated with sulphonylurea and found a non-significant decrease in heart

failure for those receiving metformin.

Results for coronary heart disease also could not be meta-analysed. One study (Peters et al.,

2013) found a non-significant reduction for users of metformin compared to other diabetics

who were not receiving metformin, while another (Sullivan et al., 2011) found significant

16

increases with metformin use compared to patients being treated with diet alone or

sulphonylurea monotherapy. Other CVD outcomes investigated were left ventricular

dysfunction, which was significantly more prevalent in diabetics being treated with

metformin compared to those not taking metformin (OR=1.62, 95%CI 1.09 to 2.40, p=0.16) –

although the authors of this cross-sectional study concluded that this outcome was likely

due to confounding due to numerous large between group differences (Giorda et al., 2011) –

and macrovascular morbidity and mortality (Kooy et al., 2009) which was significantly

lowered by the addition of metformin to insulin therapy compared to the addition of a

placebo to insulin therapy (HR= 0.34, 95%CI 0.21 to 0.56, p=0.001).

3.2.3 Other diseases of ageing

Seven studies investigated the effect of metformin on diseases of ageing, other than cancer

and CVD, although none could be combined in meta-analysis (Supp table 3). Critical appraisal

of studies (supplementary material 2) showed differences between groups at baseline as

well as short follow up with no reasons given for losses. Two studies compared outcomes for

people using metformin to the general or non-diabetic population; neither reported on

differences in ages, although one was a case-control study which matched for age

(Geoghegan et al., 2004) and the other adjusted for age (Vestergaard et al., 2005). Amongst

the five studies which compared people receiving metformin to those controlling their

diabetes through other means there was heterogeneity in reporting; two studies included a

cohort of metformin patients who were younger than their comparators (Hung et al., 2013;

Masica et al., 2013), one reported that one subgroup of metformin users was younger and

another older but did not report the overall comparison (Ng et al., 2014), one reported that

metformin users were slightly older than patients assigned to a different medication (in an

RCT design) (Kahn et al., 2008), and one did not report relative ages (Lin et al., 2015). All

studies adjusted for age in their analyses with the exception of the RCT. Two studies

investigated kidney failure or decline. One study (Hung et al., 2013) found a non-significant

decrease for the incidence of a glomerular filtration event or end stage renal disease for

diabetic patients treated with metformin compared to diabetic patients treated with

sulphonylurea, however this decrease was significant in a subgroup of patients where urine

protein at baseline was known and adjusted for (HR=0.78, 95%CI 0.64 to 0.97). The same

analyses were performed for patients treated with metformin in addition to sulphonylurea

compared to sulphonylurea alone, however no significant differences were found. The

second study found a non-significant reduction in the decline of estimated glomerular

17

filtration rate to <60ml/min/1.73m2 within one year for metformin users (Masica et al.,

2013) compared to sulphonylurea users, however there was a significant reduction in risk of

developing proteinuria (HR= 0.71, 95%CI 0.53 to 0.95, p<0.05). These comparisons were

repeated for treatment with thiazolidinediones compared to metformin, and no significant

differences were found.

The effect of metformin on fracture risk was investigated in two studies. One study (Kahn et

al., 2008) showed that metformin significantly decreased risk of fracture relative to

treatment with rosiglitazone (HR= 0.64, 95%CI 0.46 to 0.88, p=0.007). Stratified analysis

showed that the effect persisted in women but not in men. The other study (Vestergaard et

al., 2005) found the metformin users had lower risks of fracture compared to the general

population for <150 defined daily doses (DDD) (OR=0.87, 95%CI 0.86 to 0.96, p<0.05), 150-

499 DDD (OR= 0.81, 95%CI 0.71 to 0.94, p<0.05), and ≥500 DDD (OR= 0.81, 95%CI 0.70 to

0.93, p<0.05) consumed from the study start date to the date of censoring.

Additional studies on diseases of ageing investigated open angle glaucoma (Lin et al., 2015),

which had a significantly lower incidence in diabetics taking metformin compared to other

diabetics (HR= 0.75, 95%CI 0.59 to 0.95, p<0.05), cognitive impairment (Ng et al., 2014),

which was significantly less likely to develop in patients taking metformin compared to other

diabetics (OR= 0.49, 95%CI 0.25 to 0.95, p<0.05; a finding which remained significant in users

who took metformin for more than 6 years before baseline (OR= 0.27, 95%CI 0.12 to 0.60,

p<0.05), but not those who took it for 6 years or less), and carpal tunnel syndrome

(Geoghegan et al., 2004), which was not significantly associated with metformin use relative

to the general population.

4.0 Discussion

This systematic review has shown through meta-analysis that diabetics taking metformin

have a lower rate of all-cause mortality than non-diabetic people and the general

population. Our results suggest that metformin could be an effective intervention to extend

the lifespans of people who do not have diabetes. This is supported by additional meta-

analyses showing that diabetics taking metformin had lower rates of all-cause mortality than

18

diabetics receiving other therapies after adjusting for relative disease control. This was true

for metformin compared to diabetics receiving any non-metformin therapy, insulin, or

sulphonylurea, however, the result was non-significant for diet therapy. This finding could be

due to diet therapy being a more likely option for people with early or less severe diabetes,

or it could alternatively be explained by improved diet and lifestyle having an effect beyond

disease control comparable to metformin. With only two studies contributing to the meta-

analysis further research is needed for conclusions to be drawn.

Other results suggest that metformin’s effect on all-cause mortality could be due to it having

a geroprotective effect of delaying or preventing diseases of ageing, particularly cancer and

CVD which are two of the leading causes of death and disability worldwide (Mathers and

Loncar, 2006). We show that diabetic people taking metformin had a lower rate of

developing any cancer compared with the general population, and had a lower risk of

developing colorectal, breast or lung cancer compared with diabetics managing their

diabetes through non-metformin therapies after adjusting for disease control. The rate of

any form of CVD was similarly lower for diabetic people taking metformin compared to

those managing their diabetes through any non-metformin therapy or insulin therapy after

adjusting for disease control, however findings for sulphonylurea were non-significant. The

incidence of stroke was also reduced with metformin use compared to any non-metformin

therapy; however there was no significant effect for metformin on myocardial infarction.

These findings rest on the credibility of two assumptions: that improvements in survival and

reduced onset of diseases of ageing seen in diabetics above the outcomes of people who do

not have diabetes can be generalised to the non-diabetic population, and that benefits seen

between populations receiving different therapies for diabetes (which are still present after

adjusting for disease control) are independent on the therapies’ effects on diabetes and are

likewise generalisable. Of the two, the former is the more robust assumption, although the

potential for confounding, as with any observational research, does exist. As such, the

findings for all-cause mortality and cancer – which are backed by comparisons to the non-

diabetic population – should be considered more reliable than those for CVD – which are

based solely on comparisons to diabetics treated with other therapies. However, taken

together these results provide strong evidence, from a combined population of 417,316 for

all-cause mortality compared to the non-diabetic or general population alone, to support the

19

hypothesis that metformin has a geroprotective effect in humans, slowing the progression of

diseases of ageing and in doing so extending both health and lifespans.

Beyond the assumptions discussed, limitations of this research include that none of the

meta-analyses included sufficient studies for any investigation of publication bias, study

designs were entirely observational (with two exceptions that showed benefits for

metformin on fracture (Kahn et al., 2008) and macrovascular morbidity and mortality (Kooy

et al., 2009)), and none of the data for diseases of ageing beyond cancer and CVD could be

combined through meta-analysis despite promising findings for kidney failure, fracture risk,

open angle glaucoma and cognitive impairment. With regards to cognitive impairment, in

the conduct of this review’s search, numerous additional studies were found which

investigated the effect of metformin on measures of dementia. However, these studies were

not eligible for inclusion due to investigating average level of cognitive ability (rather than

the incidence or prevalence of impairment) or not adjusting for disease control. An

additional systematic review with broader inclusion criteria relating to dementia outcomes is

now planned (Campbell et al., 2017). Diabetes, despite being a major chronic disease which

becomes progressively more common with age, was not investigated by this review as

metformin being an intervention for diabetes was too much of a complicating factor.

However, experimental studies have directly investigated the effect of prophylactic

metformin for the prevention of diabetes (in high BMI populations) and shown that it

significantly reduced its development (Andreadis et al., 2009; Knowler et al., 2009).

The primary issue with study quality in this systematic review, highlighted by critical

appraisal, was that studies rarely compared groups that did not have meaningful differences

at the beginning of the observation period. Where diabetics receiving metformin were

compared to general or non-diabetic populations, this aspect is essentially a feature of the

study design. However, for metformin compared to other diabetic therapies it is more

concerning, as it reflects a trend for people receiving metformin to be younger with shorter

durations of disease, which has the potential to bias results towards finding a protective

effect for metformin. Critical appraisal also showed that studies were generally thorough in

adjusting for potential confounding factors, however, and a sizeable proportion of studies

utilised matching. As such, the potential impact of this source of bias is at least mitigated.

Additionally, the general pattern of findings for metformin compared to other diabetic

20

therapies being reproduced by findings for metformin compared to the general or non-

diabetic population suggests that they are real and not the result of confounding. Studies

frequently had follow up periods less than five years, which affects the likelihood of age

related outcomes having an opportunity to occur, however the use of summary measures

which incorporate time (i.e. hazard and rate ratios) will have accommodated this. Finally,

studies did not report on reasons for (or the existence of) loss to follow up which, if known,

may have revealed or precluded potential sources of bias.

Future research in this area should include controlled trials. Two existing trials on the effects

of metformin on ageing – Metformin in Longevity Study (MILES NCT02432287) and Targeting

Ageing with Metformin (TAME) (Barzilai et al., 2016) – are currently planned. The findings of

this systematic review suggest that studies on the effect of metformin in older people on the

incidence of chronic diseases (healthspan) and longevity (lifespan) over the long term are

warranted. It must be stressed, however, that any immediate use of metformin as an

intervention for ageing by the general population is not supported. Despite being a relatively

safe drug, with feared effects on lactic acidosis and kidney function emerging to be minor

(Ekstrom et al., 2012; Vasisht et al., 2010), metformin is not risk free. Although studies in

mice have generally been positive, some have shown increased mortality when taken at

higher concentrations (Martin-Montalvo et al., 2013), and others have suggested that its

effects may be gender exclusive (favouring females (Anisimov et al., 2010b)), an observation

which is supported by some subgroup analyses performed in included studies (Bodmer et al.,

2012b; Kahn et al., 2008). As such, there is a risk that people taking metformin as a

geroprotective agent will experience harm or no effect even if it is ultimately found to be

beneficial in some circumstances. Additionally, recent work by Wang et al. – who showed in

an included study that the presence of frailty significantly attenuated the effect of

metformin on mortality (Wang et al., 2014) –found that older men’s propensity for diseases

of ageing significantly interacted with metformin’s effect on the development of these

diseases (Wang et al., 2017).

The findings reported in this systematic review remain preliminary generalisations, primarily

making use of existing observational evidence collected for other purposes to investigate the

credibility of the hypothesis that the insulin sensitiser metformin may extend the health and

lifespans of people from the non-diabetic population. Differences in baseline characteristics

21

were found which had the potential to bias results both towards positive findings (where

metformin users were compared to other diabetics) as well as away from (where metformin

users were compared to non-diabetics). While they should not be overstated, the apparent

association with reductions in all-cause mortality and diseases of ageing found through

meta-analysis do support this hypothesis, and metformin should be investigated as an

intervention for ageing in future clinical trials.

Funding

This research did not receive any specific grant from funding agencies in the public,

commercial, or not-for-profit sectors.

Figure legends

Figure 1. PRISMA flow chart.

Figure 2. All-cause mortality hazard ratios in people using metformin for diabetes compared

to the general or non-diabetic population.

Figure 3. All-cause mortality hazard ratios in people using metformin for diabetes compared

people using other therapies for diabetes. 2.1.2 ‘All Non-Metformin Control’ includes studies

where the controls were any diabetic people receiving who were not being treated with

metformin (or controls which were receiving a broad range of therapies). 2.1.3 ‘Diet Control’

includes studies where controls were managing their diabetes though diet. 2.1.4 ‘Insulin

Control’ includes studies were controls were managing their diabetes using insulin. 2.1.5

‘Sulphonylurea’ includes studies where controls were managing their diabetes using

sulphonylurea.

Figure 4. A) Cancer incidence in people with diabetes taking metformin compared to the

general or non-diabetic population. B) Pancreatic cancer incidence in people with diabetes

taking metformin compared to the general or non-diabetic population.

Figure 5. A) Colorectal cancer incidence for people with diabetes taking metformin

compared to diabetics not taking metformin. B) Breast cancer incidence for people with

22

diabetes taking metformin compared to diabetics not taking metformin C) Lung cancer

incidence for people with diabetes taking metformin compared to diabetics not taking

metformin.

Figure 6. Incidence of any CVD in diabetic patients treated with metformin compared to

diabetic controls receiving other therapies. 2.5.1 ‘All Non-Metformin Control’ includes

studies where the controls were any diabetic patients who were not being treated with

metformin (or controls who were receiving a broad range of therapies). 2.5.2‘Diet Control’

includes studies where controls were managing their diabetes though diet. 2.5.3 ‘Insulin

Control’ includes studies where controls were managing their diabetes using insulin 2.5.4

‘Sulphonylurea control’ includes studies where controls were managing their diabetes using

sulphonylurea.

Figure 7. A) Stroke incidence for diabetics taking metformin compared to diabetics not

taking metformin. B) Myocardial infarction incidence for diabetics taking metformin

compared to diabetics not taking metformin.

References

Algire, C., Moiseeva, O., Deschenes-Simard, X., Amrein, L., Petruccelli, L., Birman,

E., Viollet, B., Ferbeyre, G., Pollak, M.N., 2012. Metformin reduces

endogenous reactive oxygen species and associated DNA damage. Cancer

Prev Res (Phila) 5, 536-543.

Andersson, C., Vaag, A., Selmer, C., Schmiegelow, M., Sørensen, R., Lindhardsen, J.,

Gislason, G.H., Køber, L., Torp-Pedersen, C., 2012. Risk of cancer in patients

using glucose-lowering agents: A nationwide cohort study of 3.6 million

people. BMJ Open 2.

Andreadis, E.A., Katsanou, P.M., Georgiopoulos, D.X., Tsourous, G.I., Yfanti, G.K.,

Gouveri, E.T., Diamantopoulos, E.J., 2009. The effect of metformin on the

incidence of type 2 diabetes mellitus and cardiovascular disease risk factors in

overweight and obese subjects--the Carmos study. Exp Clin Endocrinol

Diabetes 117, 175-180.

Anisimov, V.N., Berstein, L.M., Popovich, I.G., Zabezhinski, M.A., Egormin, P.A.,

Piskunova, T.S., Semenchenko, A.V., Tyndyk, M.L., Yurova, M.N.,

Kovalenko, I.G., Poroshina, T.E., 2011. If started early in life, metformin

treatment increases life span and postpones tumors in female SHR mice.

Aging (Albany NY) 3, 148-157.

Anisimov, V.N., Egormin, P.A., Piskunova, T.S., Popovich, I.G., Tyndyk, M.L.,

Yurova, M.N., Zabezhinski, M.A., Anikin, I.V., Karkach, A.S., Romanyukha,

A.A., 2010a. Metformin extends life span of HER-2/neu transgenic mice and

in combination with melatonin inhibits growth of transplantable tumors in

vivo. Cell Cycle 9, 188-197.

Anisimov, V.N., Piskunova, T.S., Popovich, I.G., Zabezhinski, M.A., Tyndyk, M.L.,

Egormin, P.A., Yurova, M.V., Rosenfeld, S.V., Semenchenko, A.V.,

23

Kovalenko, I.G., Poroshina, T.E., Berstein, L.M., 2010b. Gender differences

in metformin effect on aging, life span and spontaneous tumorigenesis in

129/Sv mice. Aging (Albany NY) 2, 945-958.

Azoulay, L., Dell'Aniello, S., Gagnon, B., Pollak, M., Suissa, S., 2011. Metformin

and the incidence of prostate cancer in patients with type 2 diabetes. Cancer

Epidemiol Biomarkers Prev 20, 337-344.

Bannister, C.A., Holden, S.E., Jenkins-Jones, S., Morgan, C.L., Halcox, J.P.,

Schernthaner, G., Mukherjee, J., Currie, C.J., 2014. Can people with type 2

diabetes live longer than those without? A comparison of mortality in people

initiated with metformin or sulphonylurea monotherapy and matched, non-

diabetic controls. Diabetes Obes Metab 16, 1165-1173.

Barzilai, N., Crandall, J.P., Kritchevsky, S.B., Espeland, M.A., 2016. Metformin as a

Tool to Target Aging. Cell Metab 23, 1060-1065.

Batandier, C., Guigas, B., Detaille, D., El-Mir, M.Y., Fontaine, E., Rigoulet, M.,

Leverve, X.M., 2006. The ROS production induced by a reverse-electron flux

at respiratory-chain complex 1 is hampered by metformin. J Bioenerg

Biomembr 38, 33-42.

Baur, D.M., Klotsche, J., Hamnvik, O.P., Sievers, C., Pieper, L., Wittchen, H.U.,

Stalla, G.K., Schmid, R.M., Kales, S.N., Mantzoros, C.S., 2011. Type 2

diabetes mellitus and medications for type 2 diabetes mellitus are associated

with risk for and mortality from cancer in a German primary care cohort.

Metabolism 60, 1363-1371.

Becker, C., Jick, S.S., Meier, C.R., Bodmer, M., 2014. Metformin and the risk of head

and neck cancer: a case-control analysis. Diabetes Obes Metab 16, 1148-1154.

Becker, C., Jick, S.S., Meier, C.R., Bodmer, M., 2016. Metformin and the risk of

renal cell carcinoma: a case-control analysis. Eur J Cancer Prev.

Becker, C., Meier, C.R., Jick, S.S., Bodmer, M., 2013. Case-control analysis on

metformin and cancer of the esophagus. Cancer Causes Control 24, 1763-

1770.

Berard, E., Bongard, V., Arveiler, D., Dallongeville, J., Wagner, A., Montaye, M.,

Ruidavets, J.B., Ferrieres, J., 2011. 14-year risk of all-cause mortality

according to hypoglycemic drug exposure in a general population. European

Heart Journal 32, 973-974.

Bo, S., Ciccone, G., Rosato, R., Villois, P., Appendino, G., Ghigo, E., Grassi, G.,

2012. Cancer mortality reduction and metformin: A retrospective cohort study

in type 2 diabetic patients. Diabetes, Obesity and Metabolism 14, 23-29.

Bodmer, M., Becker, C., Jick, S.S., Meier, C.R., 2012a. Metformin does not alter the

risk of lung cancer: a case-control analysis. Lung Cancer 78, 133-137.

Bodmer, M., Becker, C., Meier, C., Jick, S.S., Meier, C.R., 2012b. Use of antidiabetic

agents and the risk of pancreatic cancer: A case-control analysis. American

Journal of Gastroenterology 107, 620-626.

Bodmer, M., Meier, C., Krahenbuhl, S., Jick, S.S., Meier, C.R., 2010. Long-term

metformin use is associated with decreased risk of breast cancer. Diabetes

Care 33, 1304-1308.

Bosco, J.L.F., Antonsen, S., Sorensen, H.T., Pedersen, L., Lash, T.L., 2011.

Metformin and incident breast cancer among diabetic women: A population-

based case-control study in Denmark. Cancer Epidemiology Biomarkers and

Prevention 20, 101-111.

24

But, A., Wang, H., Mannisto, S., Pukkala, E., Haukka, J., 2014. Assessing the effect

of treatment duration on the association between anti-diabetic medication and

cancer risk. PLoS One 9, e113162.

Cabreiro, F., Au, C., Leung, K.Y., Vergara-Irigaray, N., Cocheme, H.M., Noori, T.,

Weinkove, D., Schuster, E., Greene, N.D., Gems, D., 2013. Metformin retards

aging in C. elegans by altering microbial folate and methionine metabolism.

Cell 153, 228-239.

Campbell, J.M., Stephenson, M.D., de Courten, B., Chapmean, I., Bellman, S.M.,

Aromataris, E., 2017. Metformin and Alzheimer’s disease, dementia and

cognitive impairment: a systematic review protocol. JBI Database System Rev

Implement Rep.

Chen, H.P., Shieh, J.J., Chang, C.C., Chen, T.T., Lin, J.T., Wu, M.S., Lin, J.H., Wu,

C.Y., 2013. Metformin decreases hepatocellular carcinoma risk in a dose-

dependent manner: population-based and in vitro studies. Gut 62, 606-615.

Claesen, M., Gillard, P., De Smet, F., Callens, M., De Moor, B., Mathieu, C., 2016.

Mortality in Individuals Treated With Glucose-Lowering Agents: A Large,

Controlled Cohort Study. J Clin Endocrinol Metab 101, 461-469.

Coyle, C., Cafferty, F.H., Vale, C., Langley, R.E., 2016. Metformin as an adjuvant

treatment for cancer: a systematic review and meta-analysis. Ann Oncol 27,

2184-2195.

Ekstrom, N., Schioler, L., Svensson, A.M., Eeg-Olofsson, K., Miao Jonasson, J.,

Zethelius, B., Cederholm, J., Eliasson, B., Gudbjornsdottir, S., 2012.

Effectiveness and safety of metformin in 51 675 patients with type 2 diabetes

and different levels of renal function: a cohort study from the Swedish

National Diabetes Register. BMJ Open 2.

Evans, J.M., Ogston, S.A., Emslie-Smith, A., Morris, A.D., 2006. Risk of mortality

and adverse cardiovascular outcomes in type 2 diabetes: a comparison of

patients treated with sulfonylureas and metformin. Diabetologia 49, 930-936.

Finkel, T., Holbrook, N.J., 2000. Oxidants, oxidative stress and the biology of ageing.

Nature 408, 239-247.

Floyd, J.S., Wiggins, K.L., Christiansen, M., Dublin, S., Longstreth, W.T., Smith,

N.L., McKnight, B., Heckbert, S.R., Weiss, N.S., Psaty, B.M., 2016. Case-

control study of oral glucose-lowering drugs in combination with long-acting

insulin and the risks of incident myocardial infarction and incident stroke.

Pharmacoepidemiol Drug Saf 25, 151-160.

Garinis, G.A., van der Horst, G.T., Vijg, J., Hoeijmakers, J.H., 2008. DNA damage

and ageing: new-age ideas for an age-old problem. Nat Cell Biol 10, 1241-

1247.

Gash, K.J., Chambers, A.C., Cotton, D.E., Williams, A.C., Thomas, M.G., 2017.

Potentiating the effects of radiotherapy in rectal cancer: the role of aspirin,

statins and metformin as adjuncts to therapy. Br J Cancer.

Gejl, M., Starup-Linde, J., Scheel-Thomsen, J., Gregersen, S., Vestergaard, P., 2015.

Risk of cardiovascular disease: The effects of diabetes and anti-diabetic drugs

- A nested case-control study. International Journal of Cardiology 178, 292-

296.

Geoghegan, J.M., Clark, D.I., Bainbridge, L.C., Smith, C., Hubbard, R., 2004. Risk

factors in carpal tunnel syndrome. J Hand Surg Br 29, 315-320.

Ghotbi, A.A., Kober, L., Finer, N., James, W.P., Sharma, A.M., Caterson, I.,

Coutinho, W., Van Gaal, L.F., Torp-Pedersen, C., Andersson, C., 2013.

Association of hypoglycemic treatment regimens with cardiovascular

25

outcomes in overweight and obese subjects with type 2 diabetes: a substudy of

the SCOUT trial. Diabetes Care 36, 3746-3753.

Gillespie, Z.E., Pickering, J., Eskiw, C.H., 2016. Better Living through Chemistry:

Caloric Restriction (CR) and CR Mimetics Alter Genome Function to Promote

Increased Health and Lifespan. Front Genet 7, 142.

Giorda, C.B., Cioffi, G., De Simone, G., Di Lenarda, A., Faggiano, P., Latini, R.,

Lucci, D., Maggioni, A.P., Tarantini, L., Velussi, M., Verdecchia, P.,

Comaschi, M., 2011. Predictors of early-stage left ventricular dysfunction in

type 2 diabetes: Results of DYDA study. European Journal of Cardiovascular

Prevention and Rehabilitation 18, 415-423.

Goossens, M.E., Buntinx, F., Zeegers, M.P., Driessen, J.H., De Bruin, M.L., De

Vries, F., 2015. Influence of metformin intake on the risk of bladder cancer in

type 2 diabetes patients. Br J Clin Pharmacol 80, 1464-1472.

Gosmanova, E.O., Canada, R.B., Mangold, T.A., Rawls, W.N., Wall, B.M., 2008.

Effect of metformin-containing antidiabetic regimens on all-cause mortality in

veterans with type 2 diabetes mellitus. Am J Med Sci 336, 241-247.

Hartung, D.M., Touchette, D.R., Bultemeier, N.C., Haxby, D.G., 2005. Risk of

hospitalization for heart failure associated with thiazolidinedione therapy: A

medicaid claims-based case-control study. Pharmacotherapy 25, 1329-1336.

Hung, A.M., Roumie, C.L., Greevy, R.A., Liu, X., Grijalva, C.G., Murff, H.J.,

Griffin, M.R., 2013. Kidney function decline in metformin versus sulfonylurea

initiators: assessment of time-dependent contribution of weight, blood

pressure, and glycemic control. Pharmacoepidemiol Drug Saf 22, 623-631.

Jansson, S.P., Svardsudd, K., Andersson, D.K., 2014. Effects of fasting blood glucose

levels and blood pressure and treatment of diabetes and hypertension on the

incidence of cardiovascular disease: a study of 740 patients with incident Type

2 diabetes with up to 30 years' follow-up. Diabet Med 31, 1055-1063.

Kahler, K.H., Rajan, M., Rhoads, G.G., Safford, M.M., Demissie, K., Lu, S.E.,

Pogach, L.M., 2007. Impact of oral antihyperglycemic therapy on all-cause

mortality among patients with diabetes in the Veterans Health Administration.

Diabetes Care 30, 1689-1693.

Kahn, S.E., Zinman, B., Lachin, J.M., Haffner, S.M., Herman, W.H., Holman, R.R.,

Kravitz, B.G., Yu, D., Heise, M.A., Aftring, R.P., Viberti, G., 2008.

Rosiglitazone-associated fractures in type 2 diabetes: an Analysis from A

Diabetes Outcome Progression Trial (ADOPT). Diabetes Care 31, 845-851.

Kickstein, E., Krauss, S., Thornhill, P., Rutschow, D., Zeller, R., Sharkey, J.,

Williamson, R., Fuchs, M., Kohler, A., Glossmann, H., Schneider, R.,

Sutherland, C., Schweiger, S., 2010. Biguanide metformin acts on tau

phosphorylation via mTOR/protein phosphatase 2A (PP2A) signaling. Proc

Natl Acad Sci U S A 107, 21830-21835.

Knowler, W.C., Fowler, S.E., Hamman, R.F., Christophi, C.A., Hoffman, H.J.,

Brenneman, A.T., Brown-Friday, J.O., Goldberg, R., Venditti, E., Nathan,

D.M., 2009. 10-year follow-up of diabetes incidence and weight loss in the

Diabetes Prevention Program Outcomes Study. Lancet 374, 1677-1686.

Kooy, A., De Jager, J., Lehert, P., Bets, D., Wulffelé, M.G., Donker, A.J.M.,

Stehouwer, C.D.A., 2009. Long-term effects of metformin on metabolism and

microvascular and macrovascular disease in patients with type 2 diabetes

mellitus. Archives of Internal Medicine 169, 616-625.

26

Koro, C.E., Bowlin, S.J., Weiss, S.R., 2005. Antidiabetic therapy and the risk of heart

failure in type 2 diabetic patients: An independent effect or confounding by

indication. Pharmacoepidemiology and Drug Safety 14, 697-703.

Lee, S.H., Min, K.J., 2013. Caloric restriction and its mimetics. BMB Rep 46, 181-

187.

Li, Y., Hu, L., Xia, Q., Yuan, Y., Mi, Y., 2017. The impact of metformin use on

survival in kidney cancer patients with diabetes: a meta-analysis. Int Urol

Nephrol 49, 975-981.

Libby, G., Donnelly, L.A., Donnan, P.T., Alessi, D.R., Morris, A.D., Evans, J.M.,

2009. New users of metformin are at low risk of incident cancer: a cohort

study among people with type 2 diabetes. Diabetes Care 32, 1620-1625.

Lin, H.C., Stein, J.D., Nan, B., Childers, D., Newman-Casey, P.A., Thompson, D.A.,

Richards, J.E., 2015. Association of Geroprotective Effects of Metformin and

Risk of Open-Angle Glaucoma in Persons With Diabetes Mellitus. JAMA

Ophthalmol 133, 915-923.

Liu, B., Fan, Z., Edgerton, S.M., Yang, X., Lind, S.E., Thor, A.D., 2011. Potent anti-

proliferative effects of metformin on trastuzumab-resistant breast cancer cells

via inhibition of erbB2/IGF-1 receptor interactions. Cell Cycle 10, 2959-2966.

Liu, F., Yan, L., Wang, Z., Lu, Y., Chu, Y., Li, X., Liu, Y., Rui, D., Nie, S., Xiang,

H., 2017. Metformin therapy and risk of colorectal adenomas and colorectal

cancer in type 2 diabetes mellitus patients: A systematic review and meta-

analysis. Oncotarget 8, 16017-16026.

Lu, Y., Garcia Rodriguez, L.A., Malgerud, L., Gonzalez-Perez, A., Martin-Perez, M.,

Lagergren, J., Bexelius, T.S., 2015. New-onset type 2 diabetes, elevated

HbA1c, anti-diabetic medications, and risk of pancreatic cancer. Br J Cancer

113, 1607-1614.

Ma, S., Zheng, Y., Xiao, Y., Zhou, P., Tan, H., 2017. Meta-analysis of studies using

metformin as a reducer for liver cancer risk in diabetic patients. Medicine

(Baltimore) 96, e6888.

Martin-Montalvo, A., Mercken, E.M., Mitchell, S.J., Palacios, H.H., Mote, P.L.,

Scheibye-Knudsen, M., Gomes, A.P., Ward, T.M., Minor, R.K., Blouin, M.J.,

Schwab, M., Pollak, M., Zhang, Y., Yu, Y., Becker, K.G., Bohr, V.A., Ingram,

D.K., Sinclair, D.A., Wolf, N.S., Spindler, S.R., Bernier, M., de Cabo, R.,

2013. Metformin improves healthspan and lifespan in mice. Nat Commun 4,

2192.

Masica, A.L., Ewen, E., Daoud, Y.A., Cheng, D., Franceschini, N., Kudyakov, R.E.,

Bowen, J.R., Brouwer, E.S., Wallace, D., Fleming, N.S., West, S.L., 2013.

Comparative effectiveness research using electronic health records: impacts of

oral antidiabetic drugs on the development of chronic kidney disease.

Pharmacoepidemiol Drug Saf 22, 413-422.

Mathers, C.D., Loncar, D., 2006. Projections of global mortality and burden of

disease from 2002 to 2030. PLoS Med 3, e442.

Mazzone, P.J., Rai, H., Beukemann, M., Xu, M., Jain, A., Sasidhar, M., 2012. The

effect of metformin and thiazolidinedione use on lung cancer in diabetics.

BMC Cancer 12.

McAlister, F.A., Eurich, D.T., Majumdar, S.R., Johnson, J.A., 2008. The risk of heart

failure in patients with type 2 diabetes treated with oral agent monotherapy.

European Journal of Heart Failure 10, 703-708.

27

Meng, F., Song, L., Wang, W., 2017. Metformin Improves Overall Survival of

Colorectal Cancer Patients with Diabetes: A Meta-Analysis. J Diabetes Res

2017, 5063239.

Mitchell, G.F., 2008. Effects of central arterial aging on the structure and function of

the peripheral vasculature: implications for end-organ damage. J Appl Physiol

(1985) 105, 1652-1660.

Na, H.J., Park, J.S., Pyo, J.H., Lee, S.H., Jeon, H.J., Kim, Y.S., Yoo, M.A., 2013.

Mechanism of metformin: inhibition of DNA damage and proliferative

activity in Drosophila midgut stem cell. Mech Ageing Dev 134, 381-390.

Nair, V., Sreevalsan, S., Basha, R., Abdelrahim, M., Abudayyeh, A., Rodrigues

Hoffman, A., Safe, S., 2014. Mechanism of metformin-dependent inhibition of

mammalian target of rapamycin (mTOR) and Ras activity in pancreatic

cancer: role of specificity protein (Sp) transcription factors. J Biol Chem 289,

27692-27701.

Nathan, D.M., Buse, J.B., Davidson, M.B., Heine, R.J., Holman, R.R., Sherwin, R.,

Zinman, B., 2006. Management of hyperglycemia in type 2 diabetes: A

consensus algorithm for the initiation and adjustment of therapy: a consensus

statement from the American Diabetes Association and the European

Association for the Study of Diabetes. Diabetes Care 29, 1963-1972.

Ng, T.P., Feng, L., Yap, K.B., Lee, T.S., Tan, C.H., Winblad, B., 2014. Long-term

metformin usage and cognitive function among older adults with diabetes. J

Alzheimers Dis 41, 61-68.

Nichols, G.A., Koro, C.E., Gullion, C.M., Ephross, S.A., Brown, J.B., 2005. The

incidence of congestive heart failure associated with antidiabetic therapies.

Diabetes/Metabolism Research and Reviews 21, 51-57.

Nordström, T., Clements, M., Karlsson, R., Adolfsson, J., Grönberg, H., 2015. The

risk of prostate cancer for men on aspirin, statin or antidiabetic medications.

European Journal of Cancer 51, 725-733.

Peters, K.E., Chubb, S.A., Davis, W.A., Davis, T.M., 2013. The relationship between

hypomagnesemia, metformin therapy and cardiovascular disease complicating

type 2 diabetes: the Fremantle Diabetes Study. PLoS One 8, e74355.

Pladevall, M., Riera-Guardia, N., Margulis, A.V., Varas-Lorenzo, C., Calingaert, B.,

Perez-Gutthann, S., 2016. Cardiovascular risk associated with the use of

glitazones, metformin and sufonylureas: meta-analysis of published

observational studies. BMC Cardiovasc Disord 16, 14.

Pratipanawatr, T., Rawdaree, P., Chetthakul, T., Bunnag, P., Ngarmukos, C.,

Benjasuratwong, Y., Leelawatana, R., Kosachunhanun, N., Plengvidhya, N.,

Deerochanawong, C., Suwanwalaikorn, S., Krittiyawong, S., Mongkolsomlit,

S., Komoltri, C., 2010. Thailand Diabetic Registry cohort: predicting death in

Thai diabetic patients and causes of death. J Med Assoc Thai 93 Suppl 3, S12-

20.

Redaniel, M.T., Jeffreys, M., May, M.T., Ben-Shlomo, Y., Martin, R.M., 2012.

Associations of type 2 diabetes and diabetes treatment with breast cancer risk

and mortality: a population-based cohort study among British women. Cancer

Causes Control 23, 1785-1795.

Rojas, L.B., Gomes, M.B., 2013. Metformin: an old but still the best treatment for

type 2 diabetes. Diabetol Metab Syndr 5, 6.

Roth, G.S., Ingram, D.K., 2016. Manipulation of health span and function by dietary

caloric restriction mimetics. Ann N Y Acad Sci 1363, 5-10.

28

Saisho, Y., 2015. Metformin and Inflammation: Its Potential Beyond Glucose-

lowering Effect. Endocr Metab Immune Disord Drug Targets 15, 196-205.

Sehdev, A., Shih, Y.C.T., Vekhter, B., Bissonnette, M.B., Olopade, O.I., Polite, B.N.,

2015. Metformin for primary colorectal cancer prevention in patients with

diabetes: A case-control study in a US population. Cancer 121, 1071-1078.

Smiechowski, B., Azoulay, L., Yin, H., Pollak, M.N., Suissa, S., 2013a. The use of

metformin and colorectal cancer incidence in patients with type ii diabetes

mellitus. Cancer Epidemiology Biomarkers and Prevention 22, 1877-1883.

Smiechowski, B.B., Azoulay, L., Yin, H., Pollak, M.N., Suissa, S., 2013b. The use of

metformin and the incidence of lung cancer in patients with type 2 diabetes.

Diabetes Care 36, 124-129.

Sullivan, D., Forder, P., Simes, J., Whiting, M., Kritharides, L., Merrifield, A.,

Donoghoe, M., Colman, P.G., Graham, N., Haapamaki, H., Keech, A., 2011.

Associations between the use of metformin, sulphonylureas, or diet alone and

cardiovascular outcomes in 6005 people with type 2 diabetes in the FIELD

study. Diabetes Res Clin Pract 94, 284-290.

Tang, Y.L., Zhu, L.Y., Li, Y., Yu, J., Wang, J., Zeng, X.X., Hu, K.X., Liu, J.Y., Xu,

J.X., 2017. Metformin Use Is Associated with Reduced Incidence and

Improved Survival of Endometrial Cancer: A Meta-Analysis. Biomed Res Int

2017, 5905384.

The Joanna Briggs Institute, 2014. The Joanna Briggs Institute Reviewer's Manual:

2014 edition, Australia.

Tierney, J.F., Stewart, L.A., Ghersi, D., Burdett, S., Sydes, M.R., 2007. Practical

methods for incorporating summary time-to-event data into meta-analysis.

Trials 8, 16.

Tseng, C.H., 2011. Diabetes and risk of bladder cancer: a study using the National

Health Insurance database in Taiwan. Diabetologia 54, 2009-2015.

Tseng, C.H., 2012a. Diabetes, metformin use, and colon cancer: a population-based

cohort study in Taiwan. Eur J Endocrinol 167, 409-416.

Tseng, C.H., 2012b. Thyroid cancer risk is not increased in diabetic patients. PLoS

One 7, e53096.

Vasisht, K.P., Chen, S.C., Peng, Y., Bakris, G.L., 2010. Limitations of metformin use

in patients with kidney disease: are they warranted? Diabetes Obes Metab 12,

1079-1083.

Vestergaard, P., Rejnmark, L., Mosekilde, L., 2005. Relative fracture risk in patients

with diabetes mellitus, and the impact of insulin and oral antidiabetic

medication on relative fracture risk. Diabetologia 48, 1292-1299.

Walker, E.J., Ko, A.H., Holly, E.A., Bracci, P.M., 2015. Metformin use among type 2

diabetics and risk of pancreatic cancer in a clinic-based case-control study.

International Journal of Cancer 136, E646-E653.

Wang, C.P., Lorenzo, C., Espinoza, S.E., 2014. Frailty Attenuates the Impact of

Metformin on Reducing Mortality in Older Adults with Type 2 Diabetes. J

Endocrinol Diabetes Obes 2.

Wang, C.P., Lorenzo, C., Habib, S.L., Jo, B., Espinoza, S.E., 2017. Differential

effects of metformin on age related comorbidities in older men with type 2

diabetes. J Diabetes Complications 31, 679-686.

Zheng, Z., Chen, H., Li, J., Li, T., Zheng, B., Zheng, Y., Jin, H., He, Y., Gu, Q., Xu,

X., 2012. Sirtuin 1-mediated cellular metabolic memory of high glucose via

the LKB1/AMPK/ROS pathway and therapeutic effects of metformin.

Diabetes 61, 217-228.

29

Zhou, G., Myers, R., Li, Y., Chen, Y., Shen, X., Fenyk-Melody, J., Wu, M., Ventre,

J., Doebber, T., Fujii, N., Musi, N., Hirshman, M.F., Goodyear, L.J., Moller,

D.E., 2001. Role of AMP-activated protein kinase in mechanism of metformin

action. J Clin Invest 108, 1167-1174.

Zhou, P.T., Li, B., Liu, F.R., Zhang, M.C., Wang, Q., Li, Y.Y., Xu, C., Liu, Y.H.,

Yao, Y., Li, D., 2017a. Metformin is associated with survival benefit in

pancreatic cancer patients with diabetes: a systematic review and meta-

analysis. Oncotarget 8, 25242-25250.

Zhou, X.L., Xue, W.H., Ding, X.F., Li, L.F., Dou, M.M., Zhang, W.J., Lv, Z., Fan,

Z.R., Zhao, J., Wang, L.X., 2017b. Association between metformin and the

risk of gastric cancer in patients with type 2 diabetes mellitus: a meta-analysis

of cohort studies. Oncotarget.

Figure Caption

30

Figr-1Figure 1. PRISMA flow chart.

Duplicates

n= 1,619

Papers found by search strategy

n=21,027

Title/ Abstract review

n= 19,408

Full texts reviewed

n= 260

Papers included

n= 53

Full texts excluded =207

Conference abstract N=39

Disease specific mortality N=1

No disease adjustment N=119

No relevant intervention/control N=13

No relevant outcome N=8

Retracted N=1

Review or comment N=15

Second disease N=11

Excluded by Title/ Abstract review

n= 19,148

31

Figr-2

32

Figr-3

33

Figr-4

34

Figr-5

35

Figr-6

36

Figr-7

37

38

Table 1. Characteristics of studies investigating the effect of metformin on all cause mortality

Study Exposure Comparator Population Diabetes control adjustment

N/events and Follow up

Covariates adjusted for in multivariable analysis

Metformin compared to Non-Diabetic or General Population Bannister 2014

Metformin initiation

Non diabetic matched for age, gender, same general practice, prior cancer status and smoking status

United Kingdom. Cohort. Incident diabetes and exposed to glucose lowering therapy for a minimum of 180 days (excluded if any record of secondary diabetes).

N/a Metformin= 78,241/2,663 Non-diabetic= 78,241/2,669 deaths Mean: 2.8 years

Age, modified Charlson index, gender, smoking status, prior antiplatelet therapy , prior lipid-lowering therapy, prior antihypertensive therapy, year of study index date and study arm.

Berard 2011

Metformin use at baseline

Non diabetic at baseline

France. Cohort. Middle aged men and women with diabetes. Exposure (hypoglycaemic drug use, presence of metformin) was assessed at entry only.

N/a Metformin= 40/9 Non-diabetic= 3,162/213 14 years

Duration of diabetes, history of diabetes complications, area of residence, age, gender, educational level, alcohol consumption, smoking, blood pressure, LDL and HDL cholesterol.

Bo 2012 Metformin use

General population based on age and sex

Italy. Cohort. Patients with type 2 diabetes.

N/a Metformin= NR/104 General population= NR/128 Mean= 4.5 years

None.

Claesen 2016

Metformin use at baseline (censored if discontinued for more than 9 months)

Non users of glucose lowering agents

Belgium. Cohort. Patients older than 18 prescribed metformin and non users of glucose lowering agents matched 5 to one on age, gender, cardiovascular history, associated therapies and year of start of follow up.

N/a Metformin= 42,900/3,389 Non diabetic= 214,500/16,517 5 years

Adjusted for age, associated therapies, and gender.

Metformin compared to other diabetes therapies, controlling for diabetes management

Bo 2012 Metformin use

Management of diabetes through diet alone

Italy. Cohort. Patients with type 2 diabetes.

HbA1c Metformin= 1,479/136 Diet= 620/68 Mean: 4.5 years

Adjusted for propensity score which included: propensity score which included: age, sex, diabetes duration, HbA1c, smoking, BMI, retinopathy, nephropathy, coronary or peripheral artery disease, other comorbidities, use of hypertensives, use of acetylsalicylic acid

Ekstrom 2012

Metformin use continuously for 12

Insulin use continuously for 12 months

Sweden. Cohort. Men and women aged 40 to 85 with type 2 diabetes on

HbA1c at baseline (at least 1 year after

Metformin= 14,697/1734

Age, sex, diabetes duration, HbA1c, non-HDL-C, BMI, smoking, eGFR, multidose

39

months before baseline

before baseline Or Other OHA Or Insulin and other OHA

continuous glucose lowering treatment that filed at least 3 prescriptions for their treatment (or 18 fills of multi-dose dispensed drugs) in the 12 months before baseline.

treatment commencement)

Insulin= 12,291/2,389 Other OHA= 5,171/020 Insulin and other OHA= 1,365/NR Mean: 3.9 years

dispensation, previous hospitalisation, history of CVD and CHF, microalbuminuria, and treatment with antihypertensive agents, lipid lowering agents, and cardiac glycosides

Evans 2006

Metformin monotherapy throughout study period (initiated at baseline, censored from date of switch to new therapy)

Sulfonylurea monotherapy throughout study period (initiated at baseline, censored from date of switch to new therapy)

United Kingdom. Cohort. Patients with type 2 diabetes newly prescribed OHAs.

Average HbA1c for the study period

Metformin= 2,286/107 Sulphonylurea= 3,331/597 NR

Gender, Age, Diabetes duration, previous CV hospitalisation, Smoking, mean HbA1c, mean BMI, mean systolic blood pressure, mean diastolic blood pressure, mean cholesterol, using aspirin, using statins, using beta blockers, using ACE inhibitors.

Ghotbi 2013

Metformin monotherapy Or Metformin with insulin

Insulin monotherapy OR Any other therapy (Diet, insulin, sulphonylureas) alone or in any combination

Worldwide (16 countries). Cohort. Patients who were 55 years or over with BMI 27-45 with diabetes.

Change in HbA1c values And/or HbA1c levels

Metformin= 1,631/96 Metformin with insulin= 1000/88 Insulin=1,116/138 Any other therapy= NR

Age, smoking habits, diabetes duration, congestive heart failure, history of hypertension, BMI, sex, history of CVD, tobacco use, HDL concentrations, LDL concentrations, change in HbA1c values and/or HbA1c levels, heart rate, actual systolic and diastolic blood pressure, sibutramine usage

Gosmanova 2008

Metformin (combination or montherapy)

Not receiving metformin

USA. Cohort. Veterans with type 2 diabetes receiving long term care at veterans affair medical center treated with a prescription claim for oral hypoglyemic agents

HbA1c at baseline (treatment was ongoing)

Metformin= 1,207/266, Non-metformin= 999/253 Metormin=62 months, Non-metformin= 61 months

Age, race, eGFR, HbA1c, use of insulin, ACEI/ARBs, or statins.

Kahler 2007

Metformin montherapy Metformin with sulphonylurea

Sulfonylurea monotherapy

USA. Cohort. People with diabetes at least 18 years of age who survived at least 1 year after drug assessment.

HbA1c at baseline (treatment was ongoing)

Metformin= 2,988/82 Metofrmin with sulphonylurea= 13,820/468 Sulphonylurea= 19,053/1,005 NR

Age, diabetes suration, HbA1C, creatinine, diabetes related physician visits and also a propensity score constructed from 48 variables.

40

Libby 2009

In receipt of a metformin prescription

No record of metformin use, matched by year of diagnosis

United Kingdom. Cohort. Patients diagnosed with type 2 diabetes aged 35 or over.

HbA1C during the study period

Metformin= 1,085/609 Non-metformin= 4,085/1,422 NR

Age, sex, smoking, deprivation, BMI, A1C, insulin use, sulphoylurea use.

Pratipanawatr 2010

Metformin use

Non use of metformin

Thailand. Cohort. Diabetic patients receiving clinic treatment

HbA1C at baseline (treatment was ongoing)

Overall= 9,370/424 3 years

Age, sex, HbA1c, serum creatinine, healthcare plan, education staus, smoking status, previous history of coronary artery disease and cerebrovascular disease, lipid lowering medication, insulin.

Sullivan 2011

Metformin montherapy

Diet alone Or Sulphonylurea monotherapy

International. Cohort. Patients with type 2 diabetes

HbA1c at baseline (treatment was ongoing)

Metformin= 1,746/NR Diet= 1,632 Sulphonylurea= 1,632 5 years

Age, sex, duration of diabetes, smoking, waist-hipratio, systolic blood pressure, total cholesterol, HDL cholesterol, HbA1c, ACR group, history of CVD, presence ofmicrovascular disease, creatinine and peripheral damage to feet.

Wang 2014

Metformin prescription as sole class of glucose lowering medication for ≥180 days

Sulfonylurea prescription as sole class of glucose lowering medication for ≥ 180 days

USA. Cohort. Veterans aged 65-89 years with type 2 diabetes without history of liver, renal diseases, or cancers

HbA1c during the study period

Metformin= 307/NR Sulphonylurea= 2,108 Mean=5.6 years

Age, race, diabetes duration, age adjusted Charlson co-morbidity, smoking cessation status, mean LDL across study period, mean HbA1c across study period and propensity score.

41

Supp table 1 Characteristics of studies investigating the effect of metformin on cancer

incidence

Study Exposure Comparator Population Diabetes control adjustment

N/events and Follow up

Covariates adjusted for in multivariable analysis

Metformin compared to Non-Diabetic or General Population

Andersson 2012

Metformin (duration of treatment calculated from number and strength of tablets)

Non users of glucose lowering agents (presumed non-diabetic).

Denmark. Cohort. Individuals who had never had cancer aged ≥35 years

N/a Cancer incidence Metformin= 11,359/5,450 Non-diabetic= 3,447,904/ 293,878 NR

Charlson comorbidity score, gender, age, calendar year

But 2014

Metformin users

Non-users of anti-diabetic medications (presumed non-diabetic).

Finland. Cohort. Stratified randomly drawn sample (sex, 10 year age groups, geographical area).

N/a Cancer incidence Metformin= 1,88/ 42 Non-diabetic= 22,093/ 1,029 Median 9 years

Age, gender, calendar time, BMI, smoking status, interaction of age and gender, age and BMI.

Tseng 2011

Metformin use

General population

Taiwan. Cohort. People without prior bladder cancer

N/a Bladder cancer Overall= 1,000,000/589 2 years

Age, sex, diabetes, nephropathy, urinary tract disease, Sulfonylurea, Acarbose, THZ, Insulin, hypertension, COPD, Stroke, ischemic heart disease, peripheral arterial disease, eye disease, dyslipidaemia, statin, fibrate, ACE inhibitor, calcium channel blocker, living region, occupation

Tseng 2012b

Metformin use

General population

Taiwan. Cohort. People without type 1 diabetes

N/a Thyroid cancer Overall=99,739 Metformin= NR/37 Gen Pop= NR/906 NR

Age, sex, comorbidities, medications, living region, occupation, potential detection examinations

Tseng 2012a

Metformin use

General population

Taiwan. Cohort. People without type 1 diabetes

N/a Colorectal cancer Overall= 995,843

Age, sex, diabetes status,, dyslipidemia, obesity, hypertension, COPD, asthma, stroke,

42

Metformin= NR/203 Gen pop= NR/1,826

mephropathy, ischemic heart disease, peripheral arterial disease, eye disease, medications, region of residence, occupation, potential colon cancer detection examinations

Becker 2014

Any metformin use Or Short term metformin use (1-29 prescriptions) Or Long term metformin use (≥30 prescriptions)

Never use of metformin

United Kingdom. Case-control. Patients aged older than 90 with HIV alcoholism or any other malignancy were excluded. Controls were matched 6:1 for calendar time (index date), age, sex, general practice and number of years in the database.

N/a Head and neck cancer Any metformin= 914/112 Short term= 469/62 Long term= 445/50 Never use= 19,204/2,762 Mean 10.6 years

All other medications (sulphonylurea, Insulin and TZD), BMI, smoking, and diagnosis of diabetes melitus

Becker 2016

Any metformin use Or Short term metformin use (1-29 prescriptions) Or Long term metformin use (≥30 prescriptions)

Never use of metformin

United Kingdom. Case-control. Patients aged older than 90 with HIV alcoholism or any other malignancy were excluded. Controls were matched 6:1 for calendar time (index date), age, sex, general practice and number of years in the database.

N/a Renal cell carcinoma Any metformin= 1,405/ 246 Short term= 737/128 Long term= 668/118 Never use=23,139/ 5,249 Mean-11.2 years

All other medications (sulphonylurea, Insulin and TZD), BMI, smoking, alcohol consumption, hypertension and diagnosis of diabetes melitus

Becker 2013

Metformin exposure Short term (1-14 prescriptions) OR Medium term (15-29 prescriptions) Or Long term (≥30 prescriptions)

Never use of metformin

Patients aged 40 to 89 excluding those with HIV, alcoholism, or any malignancy prior to the index date. Controls matched for calendar time, age, sex, general practice, number of years active history with up to 10 controls.

N/a Esophageal cancer Short term= 624/63 Medium term use= 428/43 Long term use= 831/92 Never use= 40,126/3,621 Median 11.6 years

Other medications (Sulphonylureas, Insulin, TZD), BMI, and smoking.

Bodmer 2012a

Metformin exposure Short term (1-14 prescriptions) OR Medium term (15-29

No use of metformin

United Kingdom. Case-control. Patients aged 40 to 89 excluding those with HIV, alcoholism, or any malignancy prior to the index date. Controls were matched 6:1 for

N/a Lung cancer Short term= 1,346/180 Medium term= 1,225/176 Long term= 1,098/148

Other drug use (Sulphonylureas, insulin), BMI and smoking.

43

prescriptions) Or Long term (≥30 prescriptions)

calendar time, age, sex, general practice, and number of years history.

Never use= 100,180/12,539 Mean 10.1 years

Bodmer 2012b

Metformin exposure Short term (1-9 prescriptions) OR Medium term (10-29 prescriptions) Or Long term (≥30 prescriptions)

No use of metformin

United Kingdom. Case-control. Participants were below the age of 90 years had no history of other cancers, alcoholism or HIV. Controls matched 6:1 on age, sex, calendar time, general practice, and years of history. Adjusted for other medications (insulin of sulphonylurea), BMI, smoking, alcohol consumption and diabetes duration.

N/a Pancreatic cancer Short term= 165/35 Medium term 242/45 Long term= 319/55 Never use= 18,615/2,628 NR

Other medications (insulin of sulphonylurea), BMI, smoking, alcohol consumption and diabetes duration.

Lu 2015 Use of metformin

Non diabetic United Kingdom. Case-control. Patients without a history of cancer or prior pancreatic disease aged 20 to 79 years. Controls frequency matched for sex, age, and calendar year

N/a Pancreatic cancer Metformin= 552/108 Non-diabetic= 4,498/ 444

Age, sex, body mass index, smoking, alcohol drinking, Townsend deprivation index, calender year, GP visit 1 year before index date, and diabetes (when appropriate for antidiabetics).

Walker 2012

Ever use of metformin

Never use of metformin (non-diabetic)

USA. Case control. Patients aged 21 to 85 years. Patients with diabetes diagnosed within 1 year of index excluded. Controls were frequency matched by sex, and age.

N/a Pancreatic cancer Metformin= 113/53 Non-diabetic= 1,235/455

Age, sex, race, BMI, history of pancreatitis, alcohol, smoking, family history of PC, and diabetes medications.

Baur 2011

Metformin monotherapy Or Metformin alone or in combination

Non type 2 diabetic

Germany. Cross-sectional. Consecutive primary care patients

HbA1c (and non-diabetic comparators)

Cancer prevalence Metformin monotherapy= NR/ 3 Metformin combination= NR/ 15 Non-diabetic= NR/ 185

Adjusted for age, sex, HbA1c,smoking status and BMI

Nordstrom 2015

Metformin (any within 2 years of biopsy)

Non users of any antidiabetic medication (presumed non-diabetic)

Sweden. Cross-sectional. Men undergoing their first prostate biopsy

N/a Prostate cancer Metformin= 1,096/ NR

Adjusted for age, natural log transformed PSA concentration, PSA quotient, Charlson comorbidity index,

44

Non-diabetic= 17,478/NR Overall events= 8,403

educational level, use of anti-diabetic medication.

Metformin compared to other diabetes therapies, controlling for diabetes management

Gossens 2015

Metformin use (measured throughout the study period) Or Only metformin use Or Metformin with sulphonylurea

Use of sulphonylurea alone

United Kingdome. Cohort. Patients with at least one prescription of anti diabetic drugs and/or insulin, aged 18 years or over.

HbA1c record made before the start of each time interval in the previous year

Bladder cancer Metformin=132,960/ 461 Sulphonylurea= 32,438/124 Mean 5.3-7.6 years

Age, gender, smoking, BMI, HbA1c across study period

Libby 2009

In receipt of a metformin prescription

Matched patients (year of diagnosis) who had no record of metformin use

Denmark. Cohort. People aged 18 years or over without a history of colorectal cancer or bowel resection

HbA1C during the study period

Cancer incidence, colorectal cancer, lung cancer, breast cancer Metformin= 4,085/297, 40, 35, or 24 Non-metformin= 4,085/474, 76, 58, or 41 NR

Age, sex, smoking, deprivation, BMI, HbA1c, insulin use, sulphonylurea use

Redaniel 2012

Continuous prescription for at least 6 months of metformin monotherapy Or Metformin with sulphonylurea

Continuous prescription for at least 6 months of sulphonylurea monotherapy.

United kingdom. Cohort. Women with incident diabetes without a previous diagnosis of breast cancer

Time weighted average of HbA1c

Breast cancer Metformin= NR/151 Metformin+ Sulphonylurea=NR/88 Sulphonylurea= NR/93 NR

Age, period, region, BMI, year of diagnosis and weighted HbA1c

Azoulay 2011

Ever use of metformin (at least on prescription prior to index)

Ever use of any other antidiabetic agent.

United Kingdom. Case-control. Male patients who with incident use of at least one antidiabetic agent (not insulin) aged 40 years or greater. Controls were matched at up to 10:1 after matching for year of birth, and date of cohort entry.

HbA1c (last measurement before index date

Prostate cancer Metformin= 5,772/536 Non-metformin= 2,326/203 Mean 4.7 years

HbA1c at index, excessive alcohol use, obesity, smoking, lower urinary tract symptoms, previous cancer, previous use of NSAIDs, antihypertensives or statins, and ever use of other antidiabetics

45

Becker 2013

Metformin exposure Short term (1-14 prescriptions) OR Medium term (15-29 prescriptions) Or Long term (≥30 prescriptions)

No use of metformin

United Kingdom. Case-control. Patients aged 40 to 89 with diabetes excluding those with HIV, alcoholism, or any malignancy prior to the index date. Controls matched up to 10:1 for calendar time, age, sex, general practice, number of years active history.

HbA1c Esophageal cancer Short term= 730/62 Medium term use= 476/43 Long term use= 834/92 Never use= 2030/173 NR

Other medications (Sulphonylurea, Insulin, TZD), BMI, smoking, diabetes duration and HbA1c level.

Bodmer 2010

Metformin use Short term (1-9 prescriptions) Medium term (10-39 prescriptions) Long term (≥40 prescriptions)

No use of metformin

United Kingdom. Case-control. Diabetic women aged 30-79 years, excluding those with alcoholism or a diagnosis of gestational diabetes. Controls matched 4:1 for age, sex, general practice, and calendar time.

Last reported HbA1C

Breast cancer Short term= 269/64 Medium term= 372/84 Long term= 137/17 Never use= 680/140 NR

Sulphonylurea, THZ, Insulin, prandial glucose regulators, acarbose, estrogens, smoking, BMI, diabetes duration, and HbA1C.

Bosco 2011

Metformin use (redeemed prescriptions for at least 1 year) Or Metformin use (redeemed prescriptions for at least 5 year2)

No metformin Or Other anti diabetic medication OR Diet/exercise only

Denmark. Case-control. Women with type 2 diabetes aged 50 or over. Controls matched 10:1 by risk set sampling from the same county

Diabetes complications

Breast cancer ≥1 year metformin= 1250/96 ≥ 5 year metformin= 453/35 No metformin= 297/2803 Other medications= 2197/219 Diet/exercise= 876/78 NR

County, complications due to diabetes, clinical obesity, age at index date, postmenopausal hormone use, and multiple imputation to impute missing parity.

Chen 2013

Any use of metformin

No use of metformin

Taiwan. Case-control. People with diabetes excluding those with previous malignancies, aged younger than 20, or who had undergone previous liver surgery. Controls matched for age, gender and date of physician visit.

Physician visits per year

Hepatocellular cancer Cases with diabetes= 22,047, Controls with diabetes= 25,773 5 years

Age, gender, hepatitis B, hepatitis C, liver cirrhosis, end stage renal disease, DM duration, DM control (assessed by physician visits), use of other diabetic agents.

Mazzone 2012

Use of metformin

Non use of metformin

USA. Case-control. Patients with diabetes mellitus.

Mean HbA1c

Lung cancer Overall 1014

BMI, HbA1c, pack years of smoking.

46

(monotherapy)

Controls were matched 1:1 for sex, date of birth, and tobacco history.

Metformin= NR/127 Never use= NR/380 NR

Sehdev 2015

Metformin in the 12 months before index date

Non use of metformin

USA. Case-control. Patients aged >18 years with diabetes mellitus. Controls matched for age, sex, and geographical region

Number of hospital admissions and number of outpatient visits.

Colorectal cancer Metformin= 3,042/983 Non use= 5,004/1,699 12 months

Obesity, polycystic ovary disease, inflammatory bowel disease, sulphonylurea use, coronary artery disease, prescribed NSAIDs, insulin use, TZD use, Charlson comorbidity index, number of hospital admissions, number of outpatient visits.

Smiechowski 2013a

Ever use of metformin (at least 12 months prior to index) OR Metofmrin use <449 days Or 449-845 days Or 845-1447 days

Never use of metformin

United Kingdom. Case-control. Patients at least 40 years of age who had received at least one non-insulin antidiabetic prescription. Controls were matched 10:1 for age, sex, calendar year of cohort entry, and duration of follow up.

Most recent HbA1c

Colorectal cancer Metformin= 1,594/163 <449 days= 1,232/ 132 449-845 days= 1,197/98 845-1447 days= 1,203/ 98 Never use= 4,450/44 Mean 4.8 years

Obesity, smoking, statins, NSAIDs, aspirin, excessive alcohol use, HbA1c, diabetes duration, cholecystomy, inflammatory bowel disease, referrals to colonoscopy, referrals to sigmoidoscopy, history of polyps, previous cancer, use of sulphonylurea, THZs, insulins and other antidiabetic agents.

Smiechowski 2013b

Ever use of metformin (at least 12 months prior to index)

Never use of metformin

United Kingdome. Case-control. Patients at least 40 years of age who had received at least one non-insulin antidiabetic prescription. Controls matched 10:1 for age, sex, calendar year, and duration of follow up

Most recent HbA1c

Lung cancer Metformin= 6,611/616 Never use= 1,961/192 Mean=5.0 years

Diabetes duration, HbA1c, obesity, smoking, excessive alcohol use, previous cancer, COPD, asthma, NSAIDs, aspirin, statins, other antidiabetic drugs.

47

Supp table 2. Characteristics of studies investigating the effect of metformin on all

cardiovascular disease

Study Exposure Comparator Population Diabetes control adjustment

N/events and Follow up

Covariates adjusted for in multivariable analysis

Metformin compared to other diabetes therapies, controlling for diabetes management

Ekstrom 2012

Metformin use continuously for 12 months before baseline

Insulin use continuously for 12 months before baseline Or Other OHA Or Insulin and other OHA

Sweden. Cohort. Men and women aged 40 to 85 with type 2 diabetes on continuous glucose lowering treatment that filed at least 3 prescriptions for their treatment (or 18 fills of multi-dose dispensed drugs) in the 12 months before baseline.

HbA1c at baseline (at least 1 year after treatment commencement)

CVD (Any) Metformin= 14,317/1,734 Insulin=11,427/2,389 Other OHA=4,964/929 Insulin + Other OHA= 1,365/NR Mean 3.9 years

Age, sex, diabetes duration, HbA1c, non-HDL-C, BMI, smoking, eGFR, multidose dispensation, previous hospitalisation, history of CVD and CHF, microalbuminuria, and treatment with antihypertensive agents, lipid lowering agents, and cardiac glycosides

Evans 2006

Metformin monotherapy throughout study period (initiated at baseline, censored from date of switch to new therapy)

Sulfonylurea monotherapy throughout study period (initiated at baseline, censored from date of switch to new therapy)

United Kingdom. Cohort. Patients with type 2 diabetes newly prescribed OHAs.

Average HbA1c for the study period

CVD admission (Any) Metformin= 2,286/229; Sulfonylurea= 3,331/567 NR

Gender, Age, Diabetes duration, previous CV hospitalisation, Smoking, mean HbA1c, mean BMI, mean systolic blood pressure, mean diastolic blod pressure, mean cholesterol, using asprin, using statins, using beta blockers, using ACE inhibitors.

Jansson 2014

Treatment with metformin

Non users of metformin

Sweden. Cohort. Patients with incident diabetes during the study period

Blood glucose level (annual mean values)

CVD (Any) Metformin=189/NR Non-metformin= 551/NR Myocardial infarction Overall events= 298 Stroke overall events= 565 Mean 131-126 months

Age, sex, year of diabetes diagnosis, BMI, smoking habits, blood pressure, blood glucose level, number of previous events of the outcome under study, diabetes treatment, blood pressure treatment.

McAlister 2008

Dispensation of metformin exclusively through the duration of the study

Dispensation of sulphonylurea exclusivley through the duration of the study

Canada. Cohort. Aged 30 years or over with at least 1 year of continuous coverage who did not have a diagnosis of heart failure at baseline.

Total physician visits after diabetic medication index dispensation

Heart failure Metformin monotherapy= 1,469/208 Sulphonylurea monotherapy= 4,162/77

Age, gender, modified chronic disease score, therapies known to affect heart failure occurance or strongly correlated with factors predicting heart failure, and total physician visits after diabetic medication index dispensation.

48

Mean 4.6-4.7 years

Nichols 2005

Metformin montherapy as index therapy

Sulphonylurea monotherapy as index therapy Or Insulin monotherapy

USA. Cohort. People with a diagnosis of diabetes that did not have a history of congestive heart failure.

HbA1c (at event or censoring)

Congestive heart failure Metformin=272/NR Sulphonylurea= 1,085/NR Insulin= 51 Mean 23-26 months

Age, sex, suration of diabetes, HbA1c, ischemic heart disease, hypertension, renal insufficiency

Peters 2013

Use of metformin

Non use of metformin

Australia. Cohort. Patients with type 2 diabetes who were not using insulin at baseline

Baseline HbA1c (drug use was ongoing)

Coronary heart disease Metformin= NR/332 Other OHA= 608/NR Overall events=254 CVD Overall= 429 events Mean 12.3 years

There was conditional variable selection but considered variables were age, sex, diabetes duration, BMI, waist circumference, fasting serum glucose, HbA1c, Serum magnesium, hypomagnesium, urinary magnesium, fractional excretion urinary magnesium, renal causes, serum creatinine, Alcohol consumption, diuretic therapy, digoxin therapy, magnesium supplementation, gastrointestinal problems.

Sullivan 2011

Metformin montherapy

Diet alone Or Sulfonylurea monotherapy

International. Cohort. Patients with type 2 diabetes

HbA1c at baseline (treatment was ongoing)

Coronary heart disease Metformin=1,746/ NR Diet= 2,627/ NR Sulphonylurea= 1,632/ NR 5 years

Age, sex, duration of diabetes, smoking, waist-hipratio, systolic blood pressure, total cholesterol, HDL cholesterol, HbA1c, ACR group, history of CVD, presence of microvascular disease, creatinine and peripheral damage to feet.

Ghotbi 2013

Metformin monotherapy Or Metformin + Insulin Or Metformin alone or in combination with any other therapy

Insulin monotherapy Or Any other therapy (Diet, insulin, sulfonylureas) alone or in any combination

Worldwide. Cohort. Patients who were 55 years or over with BMI 27-45 with diabetes.

Change in HbA1c values and/or HbA1c levels

CVD (Nonfatal myocardial infarction, nonfatal stroke, resucitation after cardiac arrest, and cardiovascular death) Metformin= 1,631/137 Metformin + Insulin= 1,000/129 Insulin=1,116/187 Any non-metformin

Age, smoking habits, diabetes duration, congestive heart failure, history of hypertension, BMI, sex, history of CVD, tobacco use, HDL concentrations, LDL concentrations, Change in HbA1c valueas and/or HbA1c levels, heart rate, actual systolic and diastolic blood pressure, sibutramine usage

49

therapy= NR/NR 6 years

Floyd 2016

Current use of metformin

Never use of metformin

USA. Case-control. People with diabetes aged 30-79 who were users of long lasting insulin. Controls were frequency matched on age, sex and calendar year

HbA1C Myocardial infarction Metformin= 194/83 Never use= 409/ 228 Stroke Metformin= 157/ 40 Never use= 348/159 21 years

Adjusted for HbA1C, index year, age, sex, hypertension status, smoking, prior cardiovascular disease, atrial fibrillation, nephrotic syndrome, diastolic blood pressure, body mass index, total cholesterol, serum creatinine, duration of diabetes, and the use of statins, ACE inhibitors, and regular and rapid-acting insulin

Gejl 2015

Use of metformin

Non use of metformin

Denmark. Case-control. Patients with diabetes mellitus. Controls were matched 3:1 on gender

HbA1c CVD (ischemic heart disease, heart failure or stroke) Overall= 840/189 NR

Adjusted for: previous CE, age, diabetes duration, gender, atrial fibrillation, hypertension, alcohol related diagnosis, nephropathy, retinopathy, neuropathy, peripheral artery disease, usage of antiarrhythmic drugs, vitamin K antagonists, heparin, pentasaccharide, argatroban, thrombocyte function inhibitors, acetylsalicylic acid, cyclooxygenase 2 inhibitors, nonselective cyclooxygenase inhibitors, buprenorphine, tramadol, oxycodone,morphine, codeine, fentanyl, pethidine, glucocorticoids, bisphosphonates, benzodiazepines, antipsychotics, antiepileptic drugs, statins, antidepressants, insulins, glitazones, DPP-4 inhibitors, liraglutide, exenatide, biguanide, and β cell stimulants, as well as: LDL, HDL, total cholesterol, HbA1c, creatinine and triglycerides.

Hartung 2005

Use of metformin

Use of another OHA

USA. Case-control. Patients who had been hospitalized aged 18 years or older with a diagnosis of diabetes mellitus. Controls were frequency matched 6:1 for age and sex.

Number of previous diabetes-related office visits.

Heart failure (hospitalisations) Metformin= 526/ 75 Non-users= 1,412/ 211 NR

Charlson comorbidity score and number of previous diabetes related office visits

50

Koro 2005

Prescription for metformin

No drug exposure Or Prescription for sulphonylurea

United Kingdom. Case-control. Incident diabetic patients 30 years or older. Controls matched 6:1 for age, gender, calender year and length of history.

Adjusted for micro and macro vascular complications o diabetes (retinopathy, neuropathy, nephropathy, foot ulcers, gangrene and end stage renal disease).

Congestive heart failure Metformin= 1,067/ 152 No drug exposure= 2,401/296 Sulphonylurea= 4,138/591 Mean= 3.4

Gender, index year of CHF diagnosis, duration of diabetes, angina pectoris, myocardial infarction, ischemic heart disease, peripheral vascular disease, valvular disease, hypertension, retinopathy, nephropathy, foot ulcers, gangrene, end stage renal disease.

Kooy 2009

Metformin (850mg, 1-3 times daily) with insulin therapy

Placebo, with insulin therapy

The Netherlands. RCT. Patients with type 2 diabetes mellitus, aged 30-80

Changes in HbA1c

Macrovascular morbidity or mortality Metformin= 196/ NR Placebo= 194/ NR 4.3 years

Changes in HbA1c level, daily dose of insulin, and systolic BP, in addition to age, sex, smoking and cardiovascular history

Giorda 2011

Metformin use

No metformin use

Italy. Cross-sectional. Non-institutionalised persons aged >45 years with diabetes mellitus free from symptoms and clinical signs of cardiac disease

HbA1c Left ventricular dysfunction Metformin=531/NR No metformin= 220/NR Overall events= 450

Age, waist circumference, SBP on echo, serum triglycerides, glycaemia, UACR, HbA1c, retinopathy, physical activity, claudication, treatment with insulin, glitazones, diuretics, and doxazosin, Q wave abnormalities, and a negative P wave component, gender, heart rate, LV hypertrophy, treatment with sulfonylurea

Solini 2013

Metformin alone or in combination

No use of metformin

Italy. Cross-sectional. Individuals with type 2 diabetes consecutively attending hospital based diabetes clinics

HbA1c CVD (Any) Metformin= 8,703/ NR Non metformin= 4,944/ NR

Age, age at T2DM diagnosis, male sex, smoking, T2DM duration, HbA1c, antihyperglycemic treatment, triglycerides, HDL-C, BMI, eGFR, albuminurea, and retinopathy.

51

Supp table 3. Characteristics of studies investigating the effect of metformin on other

diseases of ageing

Study Exposure Comparator Population Diabetes control adjustment

N/events and Follow up

Covariates adjusted for in multivariable analysis

Metformin compared to Non-Diabetic or General Population

Geoghegan 2004

Use of metformin

Non use of metformin

United Kingdome. Case-control. All patients

N/a Carpal tunnel syndrome Metformin= 149/ 47 Non-users= 16,806/ 3,344 NR

Matched by age, sex, general practice and suration of available data. Adjusted for mean annual consulting rates.

Vestergaard 2005

Use of metformin <150 DDD (defined daily dose) Or Use of metformin 150-499 DDD Or Use of metformin ≥ 1,300 DDD

Never use of metformin

Denmark. Case-control. Any subjects.

N/a Fracture Cases= 124,655 Controls= 373,962 NR

Matched for age and sex. Adjusted for type 1 diabetes, type 2 diabetes, insulin use, sulphonylurea use, other OAD drugs, prior fracture, anti-epileptic drugs, use of diuretics, use of anxiolytics and sedatives, neuroleptics, antidepressants, alcoholism, statins, non statin cholesterol lowering drugs, antihypertensives, myocardial infarction, stroke, days in bed in 1999, working or not, living with another person vs alone.

Metformin compared to other diabetes therapies, controlling for diabetes management

Hung 2013

Metformin use Or Metformin + Sulphonylurea

Sulphonylurea use

USA. Cohort. Aged ≥ 18 years who filled an incident oral hypoglycemic drug prescription during the study period, excluding those with severe medical conditions, poor kidney function or heart failure

HbA1c over time

Glomerular filtration rate or end stage renal disease events Metformin= 7,728/ NR Metformin + Sulphonylurea= 1,000 Sulphonylurea= 4,425/ NR Overall 700 events NR

Age, sex, baseline creatinine, baseline blood pressure, history of hypertension, history of cardiovascular disease, baseline HbA1c, baseline BMI, the use of ACEI or ARBs, diuretics, baseline number of medications, year of cohort entry, number of outpatient visits, history of hospitalisation at baseline, marital status.

Lin 2015 Metformin use (more than 110g

Non-use of metformin

USA. Cohort. Aged 40 years or older enrolled for more

HbA1c over time

Open angle glaucoma

Age, sex, race, socioeconomic factors, geographic region,

52

within a 2 year window)

than 2 consecutive years, diagnosed with diabetes within 2 years of being enrolled.

Overall= 150,016/ 5,893 Mean 52.8-63.3 months

cormorbid ocular disease, comorbid medical conditions, type of diabetes, Charlson comorbidity index, cataract surgery, retina surgery, diabetic control measured by HbA1c over time.

Masica 2013

Metformin use (ongoing prescriptions)

Sulphonylurea use (ongoing prescriptions) Or THZ use

USA. Cohort. New users of oral anti diabetic medications.

HbA1c over time (quantified using AUC)

Glomerular filtration reduction or Proteinurea Metformin= 1,314/ NR Sulphonylurea= 209/NR TZDs= 103. Overall 133 glomerular filtration events, 72 incidences of proteinurea Mean 1.5-2.0 years

Age, sex, race, HbA1c, renal function, congestive heart failure, coronary heart disease, hypertension, BMI, diabetes diagnosis year

Ng 2014 Metformin taken in the year before baseline Or Metformin taken for 6 years or less before baseline Or Metformin taken for more than 6 years before baseline

No metformin use in the year before baseline

Singapore. Cohort. People with a diagnosis of diabetes using anti-diabetic medications.

Fasting blood glucose at baseline

Cognitive impairment Metformin= 204/ NR Non-users= 161/NR Mean 3.9 years

Age, gender, education, other antidiabetic medication use, fasting blood glucose, duration of diabetes, BMI, hypertension, cardiovascular illness or stroke, other medical comorbidities, eGFR, GDS score, APOE-∊4 allele, and duration of follow up.

Kahn 2008

Initially 500 mg metformin titrated to the maximum effective daily dose (1g twice daily)

Initially 4 mg rosiglitazone titrated to the maximum effective daily dose (4mg twice daily)

USA. Randomised controlled trial. People with type 2 diabetes (diagnosed within 3 years) naive to oral hypoglycemic drugs, aged 30-75, fasting plasma glucose 126 to 180mg/dl. Exclusion criteria were clinically significant liver disease, renal impairment, a history of lactic acidosis, unstable or severe angina,

Current A1C

Fracture Metformin=1454/ 59, Rosiglitazone= 1456/ 92 Mean 3.3-4.0 years

Current values of weight, serum creatinine, hemotocrit, calcium, A1C, and waist circumference.

53

known congestive heart failure, uncontrolled hypertension, or chronic disease requiring periodic or intermittent treatment with oral or intravenous corticosteroids or continuous use of inhaled corticosteroids.